I was told by someone at Red Hat that ISCSI performance is still several magnitudes behind using the client / driver. Thanks, -Drew -----Original Message----- From: Nathan Fish <lordcirth@gmail.com> Sent: Thursday, October 24, 2019 1:27 PM To: Ryan <rswagoner@gmail.com> Cc: ceph-users <ceph-users@ceph.com> Subject: [ceph-users] Re: iSCSI write performance Are you using Erasure Coding or replication? What is your crush rule? What SSDs and CPUs? Does each OSD use 100% of a core or more when writing? On Thu, Oct 24, 2019 at 1:22 PM Ryan <rswagoner@gmail.com> wrote:

I'm in the process of testing the iscsi target feature of ceph. The cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It consists of 5 hosts with 12 SSD OSDs per host. Some basic testing moving VMs to a ceph backed datastore is only showing 60MB/s transfers. However moving these back off the datastore is fast at 200-300MB/s.

What should I be looking at to track down the write performance issue? In comparison with the Nimble Storage arrays I can see 200-300MB/s in both directions.

Thanks, Ryan _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

They are Samsung 860 EVO 2TB SSDs. The Dell R740xd servers have dual Intel Gold 6130 CPUs and dual SAS controllers with 6 SSDs each. Top shows around 20-25% of a core being used by each OSD daemon. I am using erasure coding with crush-failure-domain=host k=3 m=2. On Thu, Oct 24, 2019 at 1:37 PM Drew Weaver <drew.weaver@thenap.com> wrote:

I was told by someone at Red Hat that ISCSI performance is still several magnitudes behind using the client / driver.

Thanks, -Drew

-----Original Message----- From: Nathan Fish <lordcirth@gmail.com> Sent: Thursday, October 24, 2019 1:27 PM To: Ryan <rswagoner@gmail.com> Cc: ceph-users <ceph-users@ceph.com> Subject: [ceph-users] Re: iSCSI write performance

Are you using Erasure Coding or replication? What is your crush rule? What SSDs and CPUs? Does each OSD use 100% of a core or more when writing?

On Thu, Oct 24, 2019 at 1:22 PM Ryan <rswagoner@gmail.com> wrote:

...

I'm in the process of testing the iscsi target feature of ceph. The

cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It consists of 5 hosts with 12 SSD OSDs per host. Some basic testing moving VMs to a ceph backed datastore is only showing 60MB/s transfers. However moving these back off the datastore is fast at 200-300MB/s.

...

What should I be looking at to track down the write performance issue?

In comparison with the Nimble Storage arrays I can see 200-300MB/s in both directions.

...

Thanks, Ryan _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

I'm using CentOS 7.7.1908 with kernel 3.10.0-1062.1.2.el7.x86_64. The workload was a VMware Storage Motion from a local SSD backed datastore to the ceph backed datastore. Performance was measured using dstat on the iscsi gateway for network traffic and ceph status as this cluster is basically idle. I changed max_data_area_mb to 256 and cmdsn_depth to 128. This appears to have given a slight improvement of maybe 10MB/s. Moving VM to the ceph backed datastore io: client: 124 KiB/s rd, 76 MiB/s wr, 95 op/s rd, 1.26k op/s wr Moving VM off the ceph backed datastore io: client: 344 MiB/s rd, 625 KiB/s wr, 5.54k op/s rd, 62 op/s wr I'm going to test bonnie++ with an rbd volume mounted directly on the iscsi gateway. Also will test bonnie++ inside a VM on a ceph backed datastore. On Thu, Oct 24, 2019 at 7:15 PM Mike Christie <mchristi@redhat.com> wrote:

On 10/24/2019 12:22 PM, Ryan wrote:

...

I'm in the process of testing the iscsi target feature of ceph. The cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It consists of 5

What kernel are you using?

...

hosts with 12 SSD OSDs per host. Some basic testing moving VMs to a ceph backed datastore is only showing 60MB/s transfers. However moving these back off the datastore is fast at 200-300MB/s.

What is the workload and what are you using to measure the throughput?

If you are using fio, what arguments are you using? And, could you change the ioengine to rbd and re-run the test from the target system so we can check if rbd is slow or iscsi?

For small IOs, 60 is about right.

For 128-512K IOs you should be able to get around 300 MB/s for writes and 600 for reads.

1. Increase max_data_area_mb. This is a kernel buffer lio/tcmu uses to pass data between the kernel and tcmu-runner. The default is only 8MB.

In gwcli cd to your disk and do:

# reconfigure max_data_area_mb %N

where N is between 8 and 2048 MBs.

2. The Linux kernel target only allows 64 commands per iscsi session by default. We increase that to 128, but you can increase this to 512.

In gwcli cd to the target dir and do

reconfigure cmdsn_depth 512

3. I think ceph-iscsi and lio work better with higher queue depths so if you are using fio you want higher numjobs and/or iodepths.

...

What should I be looking at to track down the write performance issue? In comparison with the Nimble Storage arrays I can see 200-300MB/s in both directions.

Thanks, Ryan

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

Actually this may not work if moving from a local datastore to Ceph. For iSCSI xcopy, both the source and destination need to be accessible by the target such as in moving vms across Ceph datastores. So in your case, vmotion will be handled by VMWare data mover which uses 64K block sizes. On 25/10/2019 10:28, Maged Mokhtar wrote:

For vmotion speed, check "emulate_3pc" attribute on the LIO target. If 0 (default), VMWare will issue io in 64KB blocks which gives low speed. if set to 1  this will trigger VMWare to use vaai extended copy, which activates LIO's xcopy functionality which uses 512KB block sizes by default. We also bumped the xcopy block size to 4M (rbd object size) which gives around 400 MB/s vmotion speed, the same speed can also be achieved via Veeam backups.

/Maged

On 25/10/2019 06:47, Ryan wrote:

...

I'm using CentOS 7.7.1908 with kernel 3.10.0-1062.1.2.el7.x86_64. The workload was a VMware Storage Motion from a local SSD backed datastore to the ceph backed datastore. Performance was measured using dstat on the iscsi gateway for network traffic and ceph status as this cluster is basically idle.  I changed max_data_area_mb to 256 and cmdsn_depth to 128. This appears to have given a slight improvement of maybe 10MB/s.

Moving VM to the ceph backed datastore io:     client:   124 KiB/s rd, 76 MiB/s wr, 95 op/s rd, 1.26k op/s wr

Moving VM off the ceph backed datastore   io:     client:   344 MiB/s rd, 625 KiB/s wr, 5.54k op/s rd, 62 op/s wr

I'm going to test bonnie++ with an rbd volume mounted directly on the iscsi gateway. Also will test bonnie++ inside a VM on a ceph backed datastore.

On Thu, Oct 24, 2019 at 7:15 PM Mike Christie <mchristi@redhat.com <mailto:mchristi@redhat.com>> wrote:

On 10/24/2019 12:22 PM, Ryan wrote: > I'm in the process of testing the iscsi target feature of ceph. The > cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It consists of 5

What kernel are you using?

> hosts with 12 SSD OSDs per host. Some basic testing moving VMs to a ceph > backed datastore is only showing 60MB/s transfers. However moving these > back off the datastore is fast at 200-300MB/s.

What is the workload and what are you using to measure the throughput?

If you are using fio, what arguments are you using? And, could you change the ioengine to rbd and re-run the test from the target system so we can check if rbd is slow or iscsi?

For small IOs, 60 is about right.

For 128-512K IOs you should be able to get around 300 MB/s for writes and 600 for reads.

1. Increase max_data_area_mb. This is a kernel buffer lio/tcmu uses to pass data between the kernel and tcmu-runner. The default is only 8MB.

In gwcli cd to your disk and do:

# reconfigure max_data_area_mb %N

where N is between 8 and 2048 MBs.

2. The Linux kernel target only allows 64 commands per iscsi session by default. We increase that to 128, but you can increase this to 512.

In gwcli cd to the target dir and do

reconfigure cmdsn_depth 512

3. I think ceph-iscsi and lio work better with higher queue depths so if you are using fio you want higher numjobs and/or iodepths.

> > What should I be looking at to track down the write performance issue? > In comparison with the Nimble Storage arrays I can see 200-300MB/s in > both directions. > > Thanks, > Ryan > > > _______________________________________________ > ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> > To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io> >

_______________________________________________ ceph-users mailing list --ceph-users@ceph.io To unsubscribe send an email toceph-users-leave@ceph.io

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

On 10/25/2019 09:31 AM, Ryan wrote:

I'm not seeing the emulate_3pc setting under disks/rbd/diskname when

emulate_3pc is only for kernel based backends. tcmu-runner always has xcopy on.

calling info. A google search shows that SUSE Enterprise Storage has it available. I thought I had the latest packages, but maybe not. I'm using tcmu-runner 1.5.2 and ceph-iscsi 3.3. Almost all of my VMs are currently on Nimble iSCSI storage. I've actually tested from both and performance is the same. Doing the math off the ceph status does show it using 64K blocks in both cases.

Control Values - hw_max_sectors .. 1024 - max_data_area_mb .. 256 (override) - osd_op_timeout .. 30 - qfull_timeout .. 5

On Fri, Oct 25, 2019 at 4:46 AM Maged Mokhtar <mmokhtar@petasan.org <mailto:mmokhtar@petasan.org>> wrote:

Actually this may not work if moving from a local datastore to Ceph. For iSCSI xcopy, both the source and destination need to be accessible by the target such as in moving vms across Ceph datastores. So in your case, vmotion will be handled by VMWare data mover which uses 64K block sizes.

On 25/10/2019 10:28, Maged Mokhtar wrote:

...

For vmotion speed, check "emulate_3pc" attribute on the LIO target. If 0 (default), VMWare will issue io in 64KB blocks which gives low speed. if set to 1 this will trigger VMWare to use vaai extended copy, which activates LIO's xcopy functionality which uses 512KB block sizes by default. We also bumped the xcopy block size to 4M (rbd object size) which gives around 400 MB/s vmotion speed, the same speed can also be achieved via Veeam backups.

/Maged

On 25/10/2019 06:47, Ryan wrote:

...

I'm using CentOS 7.7.1908 with kernel 3.10.0-1062.1.2.el7.x86_64. The workload was a VMware Storage Motion from a local SSD backed datastore to the ceph backed datastore. Performance was measured using dstat on the iscsi gateway for network traffic and ceph status as this cluster is basically idle. I changed max_data_area_mb to 256 and cmdsn_depth to 128. This appears to have given a slight improvement of maybe 10MB/s.

Moving VM to the ceph backed datastore io: client: 124 KiB/s rd, 76 MiB/s wr, 95 op/s rd, 1.26k op/s wr

Moving VM off the ceph backed datastore io: client: 344 MiB/s rd, 625 KiB/s wr, 5.54k op/s rd, 62 op/s wr

I'm going to test bonnie++ with an rbd volume mounted directly on the iscsi gateway. Also will test bonnie++ inside a VM on a ceph backed datastore.

On Thu, Oct 24, 2019 at 7:15 PM Mike Christie <mchristi@redhat.com <mailto:mchristi@redhat.com>> wrote:

On 10/24/2019 12:22 PM, Ryan wrote: > I'm in the process of testing the iscsi target feature of ceph. The > cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It consists of 5

What kernel are you using?

> hosts with 12 SSD OSDs per host. Some basic testing moving VMs to a ceph > backed datastore is only showing 60MB/s transfers. However moving these > back off the datastore is fast at 200-300MB/s.

What is the workload and what are you using to measure the throughput?

If you are using fio, what arguments are you using? And, could you change the ioengine to rbd and re-run the test from the target system so we can check if rbd is slow or iscsi?

For small IOs, 60 is about right.

For 128-512K IOs you should be able to get around 300 MB/s for writes and 600 for reads.

1. Increase max_data_area_mb. This is a kernel buffer lio/tcmu uses to pass data between the kernel and tcmu-runner. The default is only 8MB.

In gwcli cd to your disk and do:

# reconfigure max_data_area_mb %N

where N is between 8 and 2048 MBs.

2. The Linux kernel target only allows 64 commands per iscsi session by default. We increase that to 128, but you can increase this to 512.

In gwcli cd to the target dir and do

reconfigure cmdsn_depth 512

3. I think ceph-iscsi and lio work better with higher queue depths so if you are using fio you want higher numjobs and/or iodepths.

> > What should I be looking at to track down the write performance issue? > In comparison with the Nimble Storage arrays I can see 200-300MB/s in > both directions. > > Thanks, > Ryan > > > _______________________________________________ > ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> > To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io> >

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io>

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io>

On 10/25/2019 03:25 PM, Ryan wrote:

Can you point me to the directions for the kernel mode iscsi backend. I was following these directions https://docs.ceph.com/docs/master/rbd/iscsi-target-cli/

If you just wanted to use the krbd device /dev/rbd$N and export it with iscsi from a single iscsi target, then you can use targetcli like described here: https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/htm... Instead of /dev/sdb in the example you would just use /dev/rbd$N in the section "BLOCK (Linux BLOCK devices)". Maged and SUSE also have a special module but I am not sure where to get their kernels and manuals. Check out suse and petasan's sites for that.

Thanks, Ryan

On Fri, Oct 25, 2019 at 11:29 AM Mike Christie <mchristi@redhat.com <mailto:mchristi@redhat.com>> wrote:

On 10/25/2019 09:31 AM, Ryan wrote: > I'm not seeing the emulate_3pc setting under disks/rbd/diskname when

emulate_3pc is only for kernel based backends. tcmu-runner always has xcopy on.

> calling info. A google search shows that SUSE Enterprise Storage has it > available. I thought I had the latest packages, but maybe not. I'm using > tcmu-runner 1.5.2 and ceph-iscsi 3.3. Almost all of my VMs are currently > on Nimble iSCSI storage. I've actually tested from both and performance > is the same. Doing the math off the ceph status does show it using 64K > blocks in both cases. > > Control Values > - hw_max_sectors .. 1024 > - max_data_area_mb .. 256 (override) > - osd_op_timeout .. 30 > - qfull_timeout .. 5 > > On Fri, Oct 25, 2019 at 4:46 AM Maged Mokhtar <mmokhtar@petasan.org <mailto:mmokhtar@petasan.org> > <mailto:mmokhtar@petasan.org <mailto:mmokhtar@petasan.org>>> wrote: > > Actually this may not work if moving from a local datastore to Ceph. > For iSCSI xcopy, both the source and destination need to be > accessible by the target such as in moving vms across Ceph > datastores. So in your case, vmotion will be handled by VMWare data > mover which uses 64K block sizes. > > On 25/10/2019 10:28, Maged Mokhtar wrote: >> >> For vmotion speed, check "emulate_3pc" attribute on the LIO >> target. If 0 (default), VMWare will issue io in 64KB blocks which >> gives low speed. if set to 1 this will trigger VMWare to use vaai >> extended copy, which activates LIO's xcopy functionality which >> uses 512KB block sizes by default. We also bumped the xcopy block >> size to 4M (rbd object size) which gives around 400 MB/s vmotion >> speed, the same speed can also be achieved via Veeam backups. >> >> /Maged >> >> On 25/10/2019 06:47, Ryan wrote: >>> I'm using CentOS 7.7.1908 with kernel 3.10.0-1062.1.2.el7.x86_64. >>> The workload was a VMware Storage Motion from a local SSD backed >>> datastore to the ceph backed datastore. Performance was measured >>> using dstat on the iscsi gateway for network traffic and ceph >>> status as this cluster is basically idle. I changed >>> max_data_area_mb to 256 and cmdsn_depth to 128. This appears to >>> have given a slight improvement of maybe 10MB/s. >>> >>> Moving VM to the ceph backed datastore >>> io: >>> client: 124 KiB/s rd, 76 MiB/s wr, 95 op/s rd, 1.26k op/s wr >>> >>> Moving VM off the ceph backed datastore >>> io: >>> client: 344 MiB/s rd, 625 KiB/s wr, 5.54k op/s rd, 62 op/s wr >>> >>> I'm going to test bonnie++ with an rbd volume mounted directly on >>> the iscsi gateway. Also will test bonnie++ inside a VM on a ceph >>> backed datastore. >>> >>> On Thu, Oct 24, 2019 at 7:15 PM Mike Christie >>> <mchristi@redhat.com <mailto:mchristi@redhat.com> <mailto:mchristi@redhat.com <mailto:mchristi@redhat.com>>> wrote: >>> >>> On 10/24/2019 12:22 PM, Ryan wrote: >>> > I'm in the process of testing the iscsi target feature of >>> ceph. The >>> > cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It >>> consists of 5 >>> >>> What kernel are you using? >>> >>> > hosts with 12 SSD OSDs per host. Some basic testing moving >>> VMs to a ceph >>> > backed datastore is only showing 60MB/s transfers. However >>> moving these >>> > back off the datastore is fast at 200-300MB/s. >>> >>> What is the workload and what are you using to measure the >>> throughput? >>> >>> If you are using fio, what arguments are you using? And, >>> could you >>> change the ioengine to rbd and re-run the test from the >>> target system so >>> we can check if rbd is slow or iscsi? >>> >>> For small IOs, 60 is about right. >>> >>> For 128-512K IOs you should be able to get around 300 MB/s >>> for writes >>> and 600 for reads. >>> >>> 1. Increase max_data_area_mb. This is a kernel buffer >>> lio/tcmu uses to >>> pass data between the kernel and tcmu-runner. The default is >>> only 8MB. >>> >>> In gwcli cd to your disk and do: >>> >>> # reconfigure max_data_area_mb %N >>> >>> where N is between 8 and 2048 MBs. >>> >>> 2. The Linux kernel target only allows 64 commands per iscsi >>> session by >>> default. We increase that to 128, but you can increase this >>> to 512. >>> >>> In gwcli cd to the target dir and do >>> >>> reconfigure cmdsn_depth 512 >>> >>> 3. I think ceph-iscsi and lio work better with higher queue >>> depths so if >>> you are using fio you want higher numjobs and/or iodepths. >>> >>> > >>> > What should I be looking at to track down the write >>> performance issue? >>> > In comparison with the Nimble Storage arrays I can see >>> 200-300MB/s in >>> > both directions. >>> > >>> > Thanks, >>> > Ryan >>> > >>> > >>> > _______________________________________________ >>> > ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> >>> <mailto:ceph-users@ceph.io <mailto:ceph-users@ceph.io>> >>> > To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io> >>> <mailto:ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io>> >>> > >>> >>> >>> _______________________________________________ >>> ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> <mailto:ceph-users@ceph.io <mailto:ceph-users@ceph.io>> >>> To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io> <mailto:ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io>> >> >> _______________________________________________ >> ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> <mailto:ceph-users@ceph.io <mailto:ceph-users@ceph.io>> >> To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io> <mailto:ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io>> >

On 10/24/2019 11:47 PM, Ryan wrote:

I'm using CentOS 7.7.1908 with kernel 3.10.0-1062.1.2.el7.x86_64. The workload was a VMware Storage Motion from a local SSD backed datastore

Ignore my comments. I thought you were just doing fio like tests in the vm.

to the ceph backed datastore. Performance was measured using dstat on the iscsi gateway for network traffic and ceph status as this cluster is basically idle. I changed max_data_area_mb to 256 and cmdsn_depth to 128. This appears to have given a slight improvement of maybe 10MB/s.

Moving VM to the ceph backed datastore io: client: 124 KiB/s rd, 76 MiB/s wr, 95 op/s rd, 1.26k op/s wr

Moving VM off the ceph backed datastore io: client: 344 MiB/s rd, 625 KiB/s wr, 5.54k op/s rd, 62 op/s wr

If you run esxtop while running your test what do you see for the number of commands in the iscsi LUN's queue?

I'm going to test bonnie++ with an rbd volume mounted directly on the

To try and isolate if its the iscsi or rbd, you need to run fio with the librbd io engine. We know krbd is going to be the fastest. ceph-iscsi uses librbd so it is a better baseline. If you are not familiar with fio you can just do something like: fio --group_reporting --ioengine=rbd --direct=1 --name=librbdtest --numjobs=32 --bs=512k --iodepth=128 --size=10G --rw=write --rbd=name_of_your_image -pool=name_of_pool

iscsi gateway. Also will test bonnie++ inside a VM on a ceph backed datastore.

On Thu, Oct 24, 2019 at 7:15 PM Mike Christie <mchristi@redhat.com <mailto:mchristi@redhat.com>> wrote:

On 10/24/2019 12:22 PM, Ryan wrote: > I'm in the process of testing the iscsi target feature of ceph. The > cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It consists of 5

What kernel are you using?

> hosts with 12 SSD OSDs per host. Some basic testing moving VMs to a ceph > backed datastore is only showing 60MB/s transfers. However moving these > back off the datastore is fast at 200-300MB/s.

What is the workload and what are you using to measure the throughput?

If you are using fio, what arguments are you using? And, could you change the ioengine to rbd and re-run the test from the target system so we can check if rbd is slow or iscsi?

For small IOs, 60 is about right.

For 128-512K IOs you should be able to get around 300 MB/s for writes and 600 for reads.

1. Increase max_data_area_mb. This is a kernel buffer lio/tcmu uses to pass data between the kernel and tcmu-runner. The default is only 8MB.

In gwcli cd to your disk and do:

# reconfigure max_data_area_mb %N

where N is between 8 and 2048 MBs.

2. The Linux kernel target only allows 64 commands per iscsi session by default. We increase that to 128, but you can increase this to 512.

In gwcli cd to the target dir and do

reconfigure cmdsn_depth 512

3. I think ceph-iscsi and lio work better with higher queue depths so if you are using fio you want higher numjobs and/or iodepths.

> > What should I be looking at to track down the write performance issue? > In comparison with the Nimble Storage arrays I can see 200-300MB/s in > both directions. > > Thanks, > Ryan > > > _______________________________________________ > ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> > To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io> >

Just to clarify, it is better to separate the different performance cases: 1- regular io performance ( iops / throughput ), this should be good. 2- vmotion within datastores managed by Ceph: this will be good, as xcopy will be used. 3. vmotion between Ceph datastore and an external datastore..this will be bad. This seems the case you are testing. It is bad because between 2 different storage systems (iqns are served on different targets), vaai xcopy cannot be used and vmware does its own stuff. It moves data using 64k block size, which gives low performance...to add some flavor, it does indeed use 32 threads, but unfortunately they use co-located addresses which does not work well in Ceph as they are hitting the same rbd object, which gets serialized due to pg locks, so you will not get any palatalization. Your speed will mostly be determined by a serial 64k, so with 1 ms write latency for ssd cluster, you will get around 64 MB/s..it will be slightly higher as the extra threads have some low effect. Note your esxtop does show 32 active ios under ACTV, the QUED of zero does is not the queue depth, but rather the "queued" io the ESX would suspend in case your active reaches the maximum by adapater ( 128 ). This is just to clarify, if case 3 is not your primary concern than i would forget about it and benchmark 1 and 2 if they are relevant. Else, if 3 is important, i am not sure you can do much as it is happening within vmware..maybe there could be a way to map the external iqn to be served by the same target serving the Ceph iqn then there could be a chance the xcopy could be activated..Mike would probably know if this has any chance of working :) /Maged On 25/10/2019 22:01, Ryan wrote:

esxtop is showing a queue length of 0

Storage motion to ceph DEVICE                                PATH/WORLD/PARTITION DQLEN WQLEN ACTV QUED %USD  LOAD   CMDS/s  READS/s WRITES/s MBREAD/s MBWRTN/s DAVG/cmd KAVG/cmd GAVG/cmd QAVG/cmd naa.6001405ec60d8b82342404d929fbbd03           - 128     -   32    0   25  0.25  1442.32     0.18  1440.50 0.00    89.78    21.32     0.01    21.34     0.01

Storage motion from ceph DEVICE                                PATH/WORLD/PARTITION DQLEN WQLEN ACTV QUED %USD  LOAD   CMDS/s  READS/s WRITES/s MBREAD/s MBWRTN/s DAVG/cmd KAVG/cmd GAVG/cmd QAVG/cmd naa.6001405ec60d8b82342404d929fbbd03           - 128     -   32    0   25  0.25  4065.38  4064.83     0.36 253.52     0.00     7.57     0.01     7.58     0.00

I tried using fio like you mentioned but it was hanging with [r=0KiB/s,w=0KiB/s][r=0,w=0 IOPS] and the ETA kept climbing. I ended up using rbd bench on the ceph iscsi gateway. With a 64K write workload I'm seeing 400MB/s transfers.

rbd create test --size 100G --image-feature layering rbd map test mkfs.ext4 /dev/rbd/rbd/test mount /dev/rbd/rbd/test test

rbd create testec --size 100G --image-feature layering --data-pool rbd_ec rbd map testec mkfs.ext4 /dev/rbd/rbd/testec mount /dev/rbd/rbd/testec testec

[root@ceph-iscsi1 mnt]# rbd bench --image test --io-size 64K --io-type write --io-total 10G bench  type write io_size 65536 io_threads 16 bytes 10737418240 pattern sequential   SEC       OPS   OPS/SEC   BYTES/SEC     1      6368   6377.59  417961796.64     2     12928   6462.27  423511630.71     3     19296   6420.18  420752986.78     4     26320   6585.61  431594792.67     5     33296   6662.37  436624891.04     6     40128   6754.67  442673957.25     7     46784   6765.75  443400452.26     8     53280   6809.02  446236110.93     9     60032   6739.67  441691068.73    10     66784   6698.91  439019550.77    11     73616   6690.88  438493253.66    12     80016   6654.35  436099640.00    13     85712   6485.07  425005611.11    14     91088   6202.49  406486113.46    15     96896   6021.17  394603137.62    16    102368   5741.19  376254347.24    17    107568   5501.57  360550910.38    18    113728   5603.17  367209502.58    19    120144   5820.48  381451245.32    20    126496   5917.60  387816078.53    21    132768   6089.71  399095466.00    22    139040   6306.98  413334431.09    23    145104   6276.42  411331743.63    24    151440   6256.67  410036891.68    25    157808   6261.12  410328554.98    26    163456   6140.03  402392725.36 elapsed:    26  ops:   163840  ops/sec:  6271.36  bytes/sec: 410999626.38

[root@ceph-iscsi1 mnt]# rbd bench --image testec --io-size 64K --io-type write --io-total 10G bench  type write io_size 65536 io_threads 16 bytes 10737418240 pattern sequential   SEC       OPS   OPS/SEC   BYTES/SEC     1      7392   7415.38  485974266.41     2     14464   7243.59  474715656.29     3     22000   7341.08  481104853.50     4     29408   7352.29  481839517.16     5     37296   7459.38  488857889.75     6     44864   7494.36  491150574.57     7     52848   7676.76  503104281.98     8     60784   7756.76  508347136.11     9     68608   7835.26  513491609.52    10     76784   7902.30  517885290.67    11     84544   7935.96  520091129.45    12     92432   7916.76  518832844.57    13    100064   7855.96  514848275.43    14    107040   7692.52  504136734.09    15    114320   7499.66  491497933.56    16    121744   7436.99  487390477.85    17    129664   7438.92  487517345.01    18    136704   7326.50  480149408.39    19    144960   7587.00  497221460.09    20    153264   7796.56  510955233.33    21    160832   7814.44  512126854.90 elapsed:    21  ops:   163840  ops/sec:  7659.97  bytes/sec: 502004079.43

On Fri, Oct 25, 2019 at 11:54 AM Mike Christie <mchristi@redhat.com <mailto:mchristi@redhat.com>> wrote:

On 10/24/2019 11:47 PM, Ryan wrote: > I'm using CentOS 7.7.1908 with kernel 3.10.0-1062.1.2.el7.x86_64. The > workload was a VMware Storage Motion from a local SSD backed datastore

Ignore my comments. I thought you were just doing fio like tests in the vm.

> to the ceph backed datastore. Performance was measured using dstat on > the iscsi gateway for network traffic and ceph status as this cluster is > basically idle.  I changed max_data_area_mb to 256 and cmdsn_depth to > 128. This appears to have given a slight improvement of maybe 10MB/s. > > Moving VM to the ceph backed datastore > io: >     client:   124 KiB/s rd, 76 MiB/s wr, 95 op/s rd, 1.26k op/s wr > > Moving VM off the ceph backed datastore >   io: >     client:   344 MiB/s rd, 625 KiB/s wr, 5.54k op/s rd, 62 op/s wr >

If you run esxtop while running your test what do you see for the number of commands in the iscsi LUN's queue?

> I'm going to test bonnie++ with an rbd volume mounted directly on the

To try and isolate if its the iscsi or rbd, you need to run fio with the librbd io engine. We know krbd is going to be the fastest. ceph-iscsi uses librbd so it is a better baseline. If you are not familiar with fio you can just do something like:

fio --group_reporting --ioengine=rbd --direct=1 --name=librbdtest --numjobs=32 --bs=512k --iodepth=128 --size=10G  --rw=write --rbd=name_of_your_image -pool=name_of_pool

> iscsi gateway. Also will test bonnie++ inside a VM on a ceph backed > datastore. > > On Thu, Oct 24, 2019 at 7:15 PM Mike Christie <mchristi@redhat.com <mailto:mchristi@redhat.com> > <mailto:mchristi@redhat.com <mailto:mchristi@redhat.com>>> wrote: > >     On 10/24/2019 12:22 PM, Ryan wrote: >     > I'm in the process of testing the iscsi target feature of ceph. The >     > cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It consists of 5 > >     What kernel are you using? > >     > hosts with 12 SSD OSDs per host. Some basic testing moving VMs to >     a ceph >     > backed datastore is only showing 60MB/s transfers. However moving >     these >     > back off the datastore is fast at 200-300MB/s. > >     What is the workload and what are you using to measure the throughput? > >     If you are using fio, what arguments are you using? And, could you >     change the ioengine to rbd and re-run the test from the target system so >     we can check if rbd is slow or iscsi? > >     For small IOs, 60 is about right. > >     For 128-512K IOs you should be able to get around 300 MB/s for writes >     and 600 for reads. > >     1. Increase max_data_area_mb. This is a kernel buffer lio/tcmu uses to >     pass data between the kernel and tcmu-runner. The default is only 8MB. > >     In gwcli cd to your disk and do: > >     # reconfigure max_data_area_mb %N > >     where N is between 8 and 2048 MBs. > >     2. The Linux kernel target only allows 64 commands per iscsi session by >     default. We increase that to 128, but you can increase this to 512. > >     In gwcli cd to the target dir and do > >     reconfigure cmdsn_depth 512 > >     3. I think ceph-iscsi and lio work better with higher queue depths so if >     you are using fio you want higher numjobs and/or iodepths. > >     > >     > What should I be looking at to track down the write performance issue? >     > In comparison with the Nimble Storage arrays I can see 200-300MB/s in >     > both directions. >     > >     > Thanks, >     > Ryan >     > >     > >     > _______________________________________________ >     > ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> >     <mailto:ceph-users@ceph.io <mailto:ceph-users@ceph.io>> >     > To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io> >     <mailto:ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io>> >     > >

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

On 31/10/2019 18:45, Paul Emmerich wrote:

On Fri, Oct 25, 2019 at 11:14 PM Maged Mokhtar <mmokhtar@petasan.org> wrote:

...

3. vmotion between Ceph datastore and an external datastore..this will be bad. This seems the case you are testing. It is bad because between 2 different storage systems (iqns are served on different targets), vaai xcopy cannot be used and vmware does its own stuff. It moves data using 64k block size, which gives low performance...to add some flavor, it does indeed use 32 threads, but unfortunately they use co-located addresses which does not work well in Ceph as they are hitting the same rbd object, which gets serialized due to pg locks, so you will not get any palatalization. Your speed will mostly be determined by a serial 64k, so with 1 ms write latency for ssd cluster, you will get around 64 MB/s..it will be slightly higher as the extra threads have some low effect. Yes, vmotion is the worst IO pattern ever for a sequential copy.

However, the situation you are describing can be fixed with RBD striping v2, just make Ceph switch to another object every 64kb, see https://docs.ceph.com/docs/master/dev/file-striping/

I'm not sure about the state of striping v2 support in the kernel module, last time I checked it wasn't supported. But ceph-iscsi/tcmu-runner got quite good over the past year, I don't see any point in still using the kernel data path for iscsi nowadays.

Paul

If case 3 (migration to/from Ceph to external stores) is important, then maybe it is worth trying Veeam Quick Migration which supports live migration. We have not tried it, but maybe this tool could be smarter than vmware and use larger block sizes. They seem to have a free edition, if you try it, i would recommend bumping the iSCSI MaxIoSizeKB on the ESXi side as it helps with their backup tool. I would not recommend using striping in this case, only very specific cases benefit from it..but for general purpose vm workloads, striping will probably be multiplying your queue depth and latencies with negative effects, specially with such small sizes. Ilya added striping to kernel rbd maybe in mainstream 4.17, we had sent our own patches earlier but Ilya changes were more comprehensive. For kernel rbd and librbd/tcmu, both have pros and cons, the kernel path is used by SUSE, the main issue for us is the support for persistent reservations which is required for setting up clustering on Windows. /Maged

...

Note your esxtop does show 32 active ios under ACTV, the QUED of zero does is not the queue depth, but rather the "queued" io the ESX would suspend in case your active reaches the maximum by adapater ( 128 ).

This is just to clarify, if case 3 is not your primary concern than i would forget about it and benchmark 1 and 2 if they are relevant. Else, if 3 is important, i am not sure you can do much as it is happening within vmware..maybe there could be a way to map the external iqn to be served by the same target serving the Ceph iqn then there could be a chance the xcopy could be activated..Mike would probably know if this has any chance of working :)

/Maged

On 25/10/2019 22:01, Ryan wrote:

esxtop is showing a queue length of 0

Storage motion to ceph DEVICE PATH/WORLD/PARTITION DQLEN WQLEN ACTV QUED %USD LOAD CMDS/s READS/s WRITES/s MBREAD/s MBWRTN/s DAVG/cmd KAVG/cmd GAVG/cmd QAVG/cmd naa.6001405ec60d8b82342404d929fbbd03 - 128 - 32 0 25 0.25 1442.32 0.18 1440.50 0.00 89.78 21.32 0.01 21.34 0.01

Storage motion from ceph DEVICE PATH/WORLD/PARTITION DQLEN WQLEN ACTV QUED %USD LOAD CMDS/s READS/s WRITES/s MBREAD/s MBWRTN/s DAVG/cmd KAVG/cmd GAVG/cmd QAVG/cmd naa.6001405ec60d8b82342404d929fbbd03 - 128 - 32 0 25 0.25 4065.38 4064.83 0.36 253.52 0.00 7.57 0.01 7.58 0.00

I tried using fio like you mentioned but it was hanging with [r=0KiB/s,w=0KiB/s][r=0,w=0 IOPS] and the ETA kept climbing. I ended up using rbd bench on the ceph iscsi gateway. With a 64K write workload I'm seeing 400MB/s transfers.

rbd create test --size 100G --image-feature layering rbd map test mkfs.ext4 /dev/rbd/rbd/test mount /dev/rbd/rbd/test test

rbd create testec --size 100G --image-feature layering --data-pool rbd_ec rbd map testec mkfs.ext4 /dev/rbd/rbd/testec mount /dev/rbd/rbd/testec testec

[root@ceph-iscsi1 mnt]# rbd bench --image test --io-size 64K --io-type write --io-total 10G bench type write io_size 65536 io_threads 16 bytes 10737418240 pattern sequential SEC OPS OPS/SEC BYTES/SEC 1 6368 6377.59 417961796.64 2 12928 6462.27 423511630.71 3 19296 6420.18 420752986.78 4 26320 6585.61 431594792.67 5 33296 6662.37 436624891.04 6 40128 6754.67 442673957.25 7 46784 6765.75 443400452.26 8 53280 6809.02 446236110.93 9 60032 6739.67 441691068.73 10 66784 6698.91 439019550.77 11 73616 6690.88 438493253.66 12 80016 6654.35 436099640.00 13 85712 6485.07 425005611.11 14 91088 6202.49 406486113.46 15 96896 6021.17 394603137.62 16 102368 5741.19 376254347.24 17 107568 5501.57 360550910.38 18 113728 5603.17 367209502.58 19 120144 5820.48 381451245.32 20 126496 5917.60 387816078.53 21 132768 6089.71 399095466.00 22 139040 6306.98 413334431.09 23 145104 6276.42 411331743.63 24 151440 6256.67 410036891.68 25 157808 6261.12 410328554.98 26 163456 6140.03 402392725.36 elapsed: 26 ops: 163840 ops/sec: 6271.36 bytes/sec: 410999626.38

[root@ceph-iscsi1 mnt]# rbd bench --image testec --io-size 64K --io-type write --io-total 10G bench type write io_size 65536 io_threads 16 bytes 10737418240 pattern sequential SEC OPS OPS/SEC BYTES/SEC 1 7392 7415.38 485974266.41 2 14464 7243.59 474715656.29 3 22000 7341.08 481104853.50 4 29408 7352.29 481839517.16 5 37296 7459.38 488857889.75 6 44864 7494.36 491150574.57 7 52848 7676.76 503104281.98 8 60784 7756.76 508347136.11 9 68608 7835.26 513491609.52 10 76784 7902.30 517885290.67 11 84544 7935.96 520091129.45 12 92432 7916.76 518832844.57 13 100064 7855.96 514848275.43 14 107040 7692.52 504136734.09 15 114320 7499.66 491497933.56 16 121744 7436.99 487390477.85 17 129664 7438.92 487517345.01 18 136704 7326.50 480149408.39 19 144960 7587.00 497221460.09 20 153264 7796.56 510955233.33 21 160832 7814.44 512126854.90 elapsed: 21 ops: 163840 ops/sec: 7659.97 bytes/sec: 502004079.43

On Fri, Oct 25, 2019 at 11:54 AM Mike Christie <mchristi@redhat.com> wrote:

...

On 10/24/2019 11:47 PM, Ryan wrote:

...

I'm using CentOS 7.7.1908 with kernel 3.10.0-1062.1.2.el7.x86_64. The workload was a VMware Storage Motion from a local SSD backed datastore Ignore my comments. I thought you were just doing fio like tests in the vm.

...

to the ceph backed datastore. Performance was measured using dstat on the iscsi gateway for network traffic and ceph status as this cluster is basically idle. I changed max_data_area_mb to 256 and cmdsn_depth to 128. This appears to have given a slight improvement of maybe 10MB/s.

Moving VM to the ceph backed datastore io: client: 124 KiB/s rd, 76 MiB/s wr, 95 op/s rd, 1.26k op/s wr

Moving VM off the ceph backed datastore io: client: 344 MiB/s rd, 625 KiB/s wr, 5.54k op/s rd, 62 op/s wr

If you run esxtop while running your test what do you see for the number of commands in the iscsi LUN's queue?

...

I'm going to test bonnie++ with an rbd volume mounted directly on the To try and isolate if its the iscsi or rbd, you need to run fio with the librbd io engine. We know krbd is going to be the fastest. ceph-iscsi uses librbd so it is a better baseline. If you are not familiar with fio you can just do something like:

fio --group_reporting --ioengine=rbd --direct=1 --name=librbdtest --numjobs=32 --bs=512k --iodepth=128 --size=10G --rw=write --rbd=name_of_your_image -pool=name_of_pool

...

iscsi gateway. Also will test bonnie++ inside a VM on a ceph backed datastore.

On Thu, Oct 24, 2019 at 7:15 PM Mike Christie <mchristi@redhat.com <mailto:mchristi@redhat.com>> wrote:

On 10/24/2019 12:22 PM, Ryan wrote: > I'm in the process of testing the iscsi target feature of ceph. The > cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It consists of 5

What kernel are you using?

> hosts with 12 SSD OSDs per host. Some basic testing moving VMs to a ceph > backed datastore is only showing 60MB/s transfers. However moving these > back off the datastore is fast at 200-300MB/s.

What is the workload and what are you using to measure the throughput?

If you are using fio, what arguments are you using? And, could you change the ioengine to rbd and re-run the test from the target system so we can check if rbd is slow or iscsi?

For small IOs, 60 is about right.

For 128-512K IOs you should be able to get around 300 MB/s for writes and 600 for reads.

1. Increase max_data_area_mb. This is a kernel buffer lio/tcmu uses to pass data between the kernel and tcmu-runner. The default is only 8MB.

In gwcli cd to your disk and do:

# reconfigure max_data_area_mb %N

where N is between 8 and 2048 MBs.

2. The Linux kernel target only allows 64 commands per iscsi session by default. We increase that to 128, but you can increase this to 512.

In gwcli cd to the target dir and do

reconfigure cmdsn_depth 512

3. I think ceph-iscsi and lio work better with higher queue depths so if you are using fio you want higher numjobs and/or iodepths.

> > What should I be looking at to track down the write performance issue? > In comparison with the Nimble Storage arrays I can see 200-300MB/s in > both directions. > > Thanks, > Ryan > > > _______________________________________________ > ceph-users mailing list -- ceph-users@ceph.io <mailto:ceph-users@ceph.io> > To unsubscribe send an email to ceph-users-leave@ceph.io <mailto:ceph-users-leave@ceph.io> >

---

ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

I had to remove the disk from the target host in gwcli to change max_data_area_mb. So the disk would need to be detached. For cmdsn_depth I was able to change it live. On Fri, Oct 25, 2019 at 7:50 AM Gesiel Galvão Bernardes < gesiel.bernardes@gmail.com> wrote:

Hi,

Em qui, 24 de out de 2019 às 20:16, Mike Christie <mchristi@redhat.com> escreveu:

...

On 10/24/2019 12:22 PM, Ryan wrote:

...

I'm in the process of testing the iscsi target feature of ceph. The cluster is running ceph 14.2.4 and ceph-iscsi 3.3. It consists of 5

What kernel are you using?

...

hosts with 12 SSD OSDs per host. Some basic testing moving VMs to a ceph backed datastore is only showing 60MB/s transfers. However moving these back off the datastore is fast at 200-300MB/s.

What is the workload and what are you using to measure the throughput?

If you are using fio, what arguments are you using? And, could you change the ioengine to rbd and re-run the test from the target system so we can check if rbd is slow or iscsi?

For small IOs, 60 is about right.

For 128-512K IOs you should be able to get around 300 MB/s for writes and 600 for reads.

1. Increase max_data_area_mb. This is a kernel buffer lio/tcmu uses to pass data between the kernel and tcmu-runner. The default is only 8MB.

In gwcli cd to your disk and do:

# reconfigure max_data_area_mb %N

where N is between 8 and 2048 MBs.

...

2. The Linux kernel target only allows 64 commands per iscsi session by default. We increase that to 128, but you can increase this to 512.

In gwcli cd to the target dir and do

reconfigure cmdsn_depth 512

For these commands, need the disk is detached or can be running "hot"?

...

3. I think ceph-iscsi and lio work better with higher queue depths so if you are using fio you want higher numjobs and/or iodepths.

...

What should I be looking at to track down the write performance issue? In comparison with the Nimble Storage arrays I can see 200-300MB/s in both directions.

Thanks, Ryan

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io