help me understand ceph snapshot (also for gfarnum@redhat.com)
good morning, i am trying to understand how ceph snapshot works. i have read that snap are cow, which means if i am correct that if a new write update an exising block on a volume, the "old" block is copied to snap before overwrite it on original volume, am i right? so, i creted a volume say 10 GB in size, empty, then created a snap. so, coming to my doubt, why the snap is 10GB in size? it should be 0, because no new write update were done, am i right? thank you VERY much for your time.
Den tors 29 feb. 2024 kl 08:48 skrev garcetto <garcetto@gmail.com>:
good morning, i am trying to understand how ceph snapshot works. i have read that snap are cow, which means if i am correct that if a new write update an exising block on a volume, the "old" block is copied to snap before overwrite it on original volume, am i right? so, i creted a volume say 10 GB in size, empty, then created a snap. so, coming to my doubt, why the snap is 10GB in size? it should be 0, because no new write update were done, am i right?
This could be dependent on how you ask about the size. The snapshot should present itself as having 10G in size, if you as a consumer of it asks how large it is. If you run something like "rbd info <name>" and "rbd du <name>" you should be able to see the differences between its apparent size and how much storage it consumes on the cluster. rbd du -p glance-images 12345678-0e1e-4d51-abcd-484604d1df0a NAME PROVISIONED USED 12345678-0e1e-4d51-abcd-484604d1df0a@snap 40GiB 40GiB 12345678-0e1e-4d51-abcd-484604d1df0a 40GiB 0B <TOTAL> 40GiB 40GiB -- May the most significant bit of your life be positive.
my problem is WHY the first snap is the SAME size of the original volume...if i DONT write/update nothing on the original volume SO the COW should dont copy nothing to the snap...so why it starts with FULL original volume size? (i have check with du and info and the snap is full used) thank you On Thu, Feb 29, 2024 at 9:39 AM Janne Johansson <icepic.dz@gmail.com> wrote:
Den tors 29 feb. 2024 kl 08:48 skrev garcetto <garcetto@gmail.com>:
good morning, i am trying to understand how ceph snapshot works. i have read that snap are cow, which means if i am correct that if a new
write update an exising block on a volume, the "old" block is copied to snap before overwrite it on original volume, am i right?
so, i creted a volume say 10 GB in size, empty, then created a snap. so, coming to my doubt, why the snap is 10GB in size? it should be 0, because no new write update were done, am i right?
This could be dependent on how you ask about the size. The snapshot should present itself as having 10G in size, if you as a consumer of it asks how large it is. If you run something like "rbd info <name>" and "rbd du <name>" you should be able to see the differences between its apparent size and how much storage it consumes on the cluster.
rbd du -p glance-images 12345678-0e1e-4d51-abcd-484604d1df0a NAME PROVISIONED USED 12345678-0e1e-4d51-abcd-484604d1df0a@snap 40GiB 40GiB 12345678-0e1e-4d51-abcd-484604d1df0a 40GiB 0B <TOTAL> 40GiB 40GiB
-- May the most significant bit of your life be positive.
Just to elaborate on Janne's earlier post using your example of a 10GB image. Assume you took a snapshot of the image then wrote 4KB after the snapshot...When you connect to the snapshot from a client machine: do you want to see a read only 10GB drive with correct sectors/filesystem formatting that you are able to mount or do you want to see a 4KB data blob that cannot be mounted or viewed from client perspective ? From a dev perspective, you may be intersected to know how much extra bytes this snapshots is taking and where/how it is stored..but that is a different view. /maged On 29/02/2024 12:44, garcetto wrote:
my problem is WHY the first snap is the SAME size of the original volume...if i DONT write/update nothing on the original volume SO the COW should dont copy nothing to the snap...so why it starts with FULL original volume size? (i have check with du and info and the snap is full used) thank you
On Thu, Feb 29, 2024 at 9:39 AM Janne Johansson <icepic.dz@gmail.com> wrote:
Den tors 29 feb. 2024 kl 08:48 skrev garcetto <garcetto@gmail.com>: > > good morning, > i am trying to understand how ceph snapshot works. > i have read that snap are cow, which means if i am correct that if a new write update an exising block on a volume, the "old" block is copied to snap before overwrite it on original volume, am i right? > so, i creted a volume say 10 GB in size, empty, then created a snap. > so, coming to my doubt, why the snap is 10GB in size? it should be 0, because no new write update were done, am i right?
This could be dependent on how you ask about the size. The snapshot should present itself as having 10G in size, if you as a consumer of it asks how large it is. If you run something like "rbd info <name>" and "rbd du <name>" you should be able to see the differences between its apparent size and how much storage it consumes on the cluster.
rbd du -p glance-images 12345678-0e1e-4d51-abcd-484604d1df0a NAME PROVISIONED USED 12345678-0e1e-4d51-abcd-484604d1df0a@snap 40GiB 40GiB 12345678-0e1e-4d51-abcd-484604d1df0a 40GiB 0B <TOTAL> 40GiB 40GiB
-- May the most significant bit of your life be positive.
_______________________________________________ Dev mailing list --dev@ceph.io To unsubscribe send an email todev-leave@ceph.io
sorry but i dont understand the client part...could you elaborate? i just want to understand why this is *before *take a snap: root@mg-ceph-90:/# rbd du volumes/volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b NAME PROVISIONED USED volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b * 1 GiB 1 GiB* root@mg-ceph-90:/# this is *after*: root@xxx:/# rbd du volumes/volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b NAME PROVISIONED USED volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b@volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b_2024-02-29T14:21:19.106+01:00 *1 GiB 1 GiB* volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b *1 GiB 0 B* <TOTAL> 1 GiB 1 GiB root@xxx:/# thank you. On Thu, Feb 29, 2024 at 1:10 PM Maged Mokhtar <mmokhtar@petasan.org> wrote:
Just to elaborate on Janne's earlier post using your example of a 10GB image. Assume you took a snapshot of the image then wrote 4KB after the snapshot...When you connect to the snapshot from a client machine: do you want to see a read only 10GB drive with correct sectors/filesystem formatting that you are able to mount or do you want to see a 4KB data blob that cannot be mounted or viewed from client perspective ?
From a dev perspective, you may be intersected to know how much extra bytes this snapshots is taking and where/how it is stored..but that is a different view.
/maged On 29/02/2024 12:44, garcetto wrote:
my problem is WHY the first snap is the SAME size of the original volume...if i DONT write/update nothing on the original volume SO the COW should dont copy nothing to the snap...so why it starts with FULL original volume size? (i have check with du and info and the snap is full used) thank you
On Thu, Feb 29, 2024 at 9:39 AM Janne Johansson <icepic.dz@gmail.com> wrote:
Den tors 29 feb. 2024 kl 08:48 skrev garcetto <garcetto@gmail.com>:
good morning, i am trying to understand how ceph snapshot works. i have read that snap are cow, which means if i am correct that if a
new write update an exising block on a volume, the "old" block is copied to snap before overwrite it on original volume, am i right?
so, i creted a volume say 10 GB in size, empty, then created a snap. so, coming to my doubt, why the snap is 10GB in size? it should be 0, because no new write update were done, am i right?
This could be dependent on how you ask about the size. The snapshot should present itself as having 10G in size, if you as a consumer of it asks how large it is. If you run something like "rbd info <name>" and "rbd du <name>" you should be able to see the differences between its apparent size and how much storage it consumes on the cluster.
rbd du -p glance-images 12345678-0e1e-4d51-abcd-484604d1df0a NAME PROVISIONED USED 12345678-0e1e-4d51-abcd-484604d1df0a@snap 40GiB 40GiB 12345678-0e1e-4d51-abcd-484604d1df0a 40GiB 0B <TOTAL> 40GiB 40GiB
-- May the most significant bit of your life be positive.
_______________________________________________ Dev mailing list -- dev@ceph.io To unsubscribe send an email to dev-leave@ceph.io
_______________________________________________ Dev mailing list -- dev@ceph.io To unsubscribe send an email to dev-leave@ceph.io
i was trying to answer your earlier question:
so, i creted a volume say 10 GB in size, empty, then created a snap. so, coming to my doubt, why the snap is 10GB in size?
On 29/02/2024 15:22, garcetto wrote:
sorry but i dont understand the client part...could you elaborate? i just want to understand why this is *_before _*take a snap:
root@mg-ceph-90:/# rbd du volumes/volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b NAME PROVISIONED USED volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b *_ 1 GiB 1 GiB_* root@mg-ceph-90:/#
this is *_after_*:
root@xxx:/# rbd du volumes/volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b NAME PROVISIONED USED volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b@volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b_2024-02-29T14:21:19.106+01:00 _*1 GiB 1 GiB*_ volume-599d5e1f-9c5a-4d2e-bd1b-c75ec2560c1b _*1 GiB 0 B*_ <TOTAL> 1 GiB 1 GiB root@xxx:/#
thank you.
On Thu, Feb 29, 2024 at 1:10 PM Maged Mokhtar <mmokhtar@petasan.org> wrote:
Just to elaborate on Janne's earlier post using your example of a 10GB image. Assume you took a snapshot of the image then wrote 4KB after the snapshot...When you connect to the snapshot from a client machine: do you want to see a read only 10GB drive with correct sectors/filesystem formatting that you are able to mount or do you want to see a 4KB data blob that cannot be mounted or viewed from client perspective ?
From a dev perspective, you may be intersected to know how much extra bytes this snapshots is taking and where/how it is stored..but that is a different view.
/maged
On 29/02/2024 12:44, garcetto wrote:
my problem is WHY the first snap is the SAME size of the original volume...if i DONT write/update nothing on the original volume SO the COW should dont copy nothing to the snap...so why it starts with FULL original volume size? (i have check with du and info and the snap is full used) thank you
On Thu, Feb 29, 2024 at 9:39 AM Janne Johansson <icepic.dz@gmail.com> wrote:
Den tors 29 feb. 2024 kl 08:48 skrev garcetto <garcetto@gmail.com>: > > good morning, > i am trying to understand how ceph snapshot works. > i have read that snap are cow, which means if i am correct that if a new write update an exising block on a volume, the "old" block is copied to snap before overwrite it on original volume, am i right? > so, i creted a volume say 10 GB in size, empty, then created a snap. > so, coming to my doubt, why the snap is 10GB in size? it should be 0, because no new write update were done, am i right?
This could be dependent on how you ask about the size. The snapshot should present itself as having 10G in size, if you as a consumer of it asks how large it is. If you run something like "rbd info <name>" and "rbd du <name>" you should be able to see the differences between its apparent size and how much storage it consumes on the cluster.
rbd du -p glance-images 12345678-0e1e-4d51-abcd-484604d1df0a NAME PROVISIONED USED 12345678-0e1e-4d51-abcd-484604d1df0a@snap 40GiB 40GiB 12345678-0e1e-4d51-abcd-484604d1df0a 40GiB 0B <TOTAL> 40GiB 40GiB
-- May the most significant bit of your life be positive.
_______________________________________________ Dev mailing list --dev@ceph.io To unsubscribe send an email todev-leave@ceph.io
_______________________________________________ Dev mailing list -- dev@ceph.io To unsubscribe send an email to dev-leave@ceph.io
_______________________________________________ Dev mailing list --dev@ceph.io To unsubscribe send an email todev-leave@ceph.io
-- Maged Mokhtar CEO PetaSAN 4 Emad El Deen Kamel Cairo 11371, Egypt www.petasan.org +201006979931 skype: maged.mokhtar
Hello, When you take a snapshot, it's not that the volume retains the data and anything that gets modified is copied to the snapshot. Rather, internally, all of the current data becomes associated with that snapshot, and reads to the volume for data that has not yet been modified are redirected to that snapshot layer. Once you write to the volume, the relevant object is copy-on-written to the volume layer. If you then take another snapshot, anything that has been copy-on-written to that point becomes associated with that new snapshot layer. HTH, Josh On Thu, Feb 29, 2024 at 3:45 AM garcetto <garcetto@gmail.com> wrote:
my problem is WHY the first snap is the SAME size of the original volume...if i DONT write/update nothing on the original volume SO the COW should dont copy nothing to the snap...so why it starts with FULL original volume size? (i have check with du and info and the snap is full used) thank you
On Thu, Feb 29, 2024 at 9:39 AM Janne Johansson <icepic.dz@gmail.com> wrote:
Den tors 29 feb. 2024 kl 08:48 skrev garcetto <garcetto@gmail.com>:
good morning, i am trying to understand how ceph snapshot works. i have read that snap are cow, which means if i am correct that if a new write update an exising block on a volume, the "old" block is copied to snap before overwrite it on original volume, am i right? so, i creted a volume say 10 GB in size, empty, then created a snap. so, coming to my doubt, why the snap is 10GB in size? it should be 0, because no new write update were done, am i right?
This could be dependent on how you ask about the size. The snapshot should present itself as having 10G in size, if you as a consumer of it asks how large it is. If you run something like "rbd info <name>" and "rbd du <name>" you should be able to see the differences between its apparent size and how much storage it consumes on the cluster.
rbd du -p glance-images 12345678-0e1e-4d51-abcd-484604d1df0a NAME PROVISIONED USED 12345678-0e1e-4d51-abcd-484604d1df0a@snap 40GiB 40GiB 12345678-0e1e-4d51-abcd-484604d1df0a 40GiB 0B <TOTAL> 40GiB 40GiB
-- May the most significant bit of your life be positive.
_______________________________________________ Dev mailing list -- dev@ceph.io To unsubscribe send an email to dev-leave@ceph.io
When you take a snapshot, it's not that the volume retains the data and anything that gets modified is copied to the snapshot.
which volume? the original or the snap? everything that is request to be written/modified first time after a snap (in COW) must be first copied to snap (original block data), then overwritten on original volume, not snap, am i right?
Rather, internally, all of the current data becomes associated with that snapshot, and reads to the volume for data that has not yet been modified are redirected to that snapshot layer.
ah ok, so read go redirected to snap, not original volume anymore? good to know thank you.
Once you write to the volume, the relevant object is copy-on-written to the volume layer. If you then take another snapshot, anything that has been copy-on-written to that point becomes associated with that new snapshot layer.
volume layer = snap right? MANY MANY THANK TO ALL FOR YOUR TIME AND PATIENCE!!! On Thu, Feb 29, 2024 at 2:21 PM Josh Baergen <jbaergen@digitalocean.com> wrote:
Hello,
When you take a snapshot, it's not that the volume retains the data and anything that gets modified is copied to the snapshot. Rather, internally, all of the current data becomes associated with that snapshot, and reads to the volume for data that has not yet been modified are redirected to that snapshot layer. Once you write to the volume, the relevant object is copy-on-written to the volume layer. If you then take another snapshot, anything that has been copy-on-written to that point becomes associated with that new snapshot layer.
HTH, Josh
On Thu, Feb 29, 2024 at 3:45 AM garcetto <garcetto@gmail.com> wrote:
my problem is WHY the first snap is the SAME size of the original
volume...if i DONT write/update nothing on the original volume SO the COW should dont copy nothing to the snap...so why it starts with FULL original volume size?
(i have check with du and info and the snap is full used) thank you
On Thu, Feb 29, 2024 at 9:39 AM Janne Johansson <icepic.dz@gmail.com> wrote:
Den tors 29 feb. 2024 kl 08:48 skrev garcetto <garcetto@gmail.com>:
good morning, i am trying to understand how ceph snapshot works. i have read that snap are cow, which means if i am correct that if a
new write update an exising block on a volume, the "old" block is copied to snap before overwrite it on original volume, am i right?
so, i creted a volume say 10 GB in size, empty, then created a snap. so, coming to my doubt, why the snap is 10GB in size? it should be 0, because no new write update were done, am i right?
This could be dependent on how you ask about the size. The snapshot should present itself as having 10G in size, if you as a consumer of it asks how large it is. If you run something like "rbd info <name>" and "rbd du <name>" you should be able to see the differences between its apparent size and how much storage it consumes on the cluster.
rbd du -p glance-images 12345678-0e1e-4d51-abcd-484604d1df0a NAME PROVISIONED USED 12345678-0e1e-4d51-abcd-484604d1df0a@snap 40GiB 40GiB 12345678-0e1e-4d51-abcd-484604d1df0a 40GiB 0B <TOTAL> 40GiB 40GiB
-- May the most significant bit of your life be positive.
_______________________________________________ Dev mailing list -- dev@ceph.io To unsubscribe send an email to dev-leave@ceph.io
participants (4)
-
garcetto
-
Janne Johansson
-
Josh Baergen
-
Maged Mokhtar