Thanks for your reply. We don't maintain it frequently. My confusion is whether the more monitor is more advantage for client(osd,rbdclient...) to get clustermap. Do All clients comunicate with  one monitor  of the  cluster at the mean time ? If not  how client to decide to communicat with which monitor? From: Nathan Fish Date: 2019-09-26 00:52 To: 展荣臻（信泰） CC: ceph-users Subject: Re: [ceph-users] how many monitor should to deploy in a 1000+ osd cluster You don't need more mons to scale; but going to 5 mons would make the cluster more robust, if it is cheap for you to do so. If you assume that 1 mon rebooting for updates or maintenance is routine, then 2/3 is vulnerable to one failure. 4/5 can survive an unexpected additional failure while one is down for maintenance. Considering your scale, this improvement in uptime might be worthwhile.   On Wed, Sep 25, 2019 at 10:26 AM 展荣臻（信泰） <zhanrzh_xt@teamsun.com.cn> wrote:

hi all:    I have a  production cluster, and it had 24 hosts (528 osds,3mons) at a former.    Now we want to add 36 hosts so the osd increase to 1320 .    does the monitor need to increase?how many numbers of monitor node is recommended?    Another question is which monitor does monclient  commnuicate with? And how it decide?    Any suggestions are welcome! _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

I’d always been led to believe that when a client attaches it tries the mons in its conf file in order until it contacts one, which may then hand it off to another. This was given as an explanation for a situation I once had where 2/5 mons were down but clients could run but couldn’t attach/boot. The idea was that since it was the *first* two mons in the clients’ config file that were down, they would time out before they could try the third. The immediate workaround was to blast out an interim config file with the first two mons commented/removed. This was in the Firefly or Hammer days, does that sound plausible and perhaps still how things work? I’ve at times bound a client (well, perhaps ‘ceph status’ isn’t a client as such?) to a specific mon with the -m switch. — aad

...

Thanks for your reply. We don't maintain it frequently. My confusion is whether the more monitor is more advantage for client(osd,rbdclient...) to get clustermap. Do All clients comunicate with one monitor of the cluster at the mean time ? If not how client to decide to communicat with which monitor?

Clients talk with one monitor, but each client to a different Monitor. They select the Monitor they talk to randomly.

Wido

...

From: Nathan Fish Date: 2019-09-26 00:52 To: 展荣臻（信泰） CC: ceph-users Subject: Re: [ceph-users] how many monitor should to deploy in a 1000+ osd cluster You don't need more mons to scale; but going to 5 mons would make the cluster more robust, if it is cheap for you to do so. If you assume that 1 mon rebooting for updates or maintenance is routine, then 2/3 is vulnerable to one failure. 4/5 can survive an unexpected additional failure while one is down for maintenance. Considering your scale, this improvement in uptime might be worthwhile.

On Wed, Sep 25, 2019 at 10:26 AM 展荣臻（信泰） <zhanrzh_xt@teamsun.com.cn> wrote:

...

hi all: I have a production cluster, and it had 24 hosts (528 osds,3mons) at a former. Now we want to add 36 hosts so the osd increase to 1320 . does the monitor need to increase?how many numbers of monitor node is recommended? Another question is which monitor does monclient commnuicate with? And how it decide? Any suggestions are welcome! _______________________________________________ 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

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