any thoughts ?
On Tue, Mar 14, 2017 at 10:22 PM, Alejandro Comisario <alejandro@xxxxxxxxxxx> wrote:
Greg, thanks for the reply.True that i cant provide enough information to know what happened since the pool is gone.But based on your experience, can i please take some of your time, and give me the TOP 5 fo what could happen / would be the reason to happen what hapened to that pool (or any pool) that makes Ceph (maybe hapened specifically in Hammer ) to behave like that ?Information that i think will be of value, is that the cluster was 5 nodes large, running "0.94.6-1trusty" i added two nodes running the latest "0.94.9-1trusty" and replication into those new disks never ended, since i saw WEIRD errors on the new OSDs, so i thought that packages needed to be the same, so i "apt-get upgraded" the 5 old nodes without restrting nothing, so rebalancing started to happen without errors (WEIRD).after these two nodes reached 100% of the disks weight, the cluster worked perfectly for about two weeks, till this happened.After the resolution from my first email, everything has been working perfect.thanks for the responses.On Fri, Mar 10, 2017 at 4:23 PM, Gregory Farnum <gfarnum@xxxxxxxxxx> wrote:On Tue, Mar 7, 2017 at 10:18 AM Alejandro Comisario <alejandro@xxxxxxxxxxx> wrote:Gregory, thanks for the response, what you've said is by far, the most enlightneen thing i know about ceph in a long time.What brings even greater doubt, which is, this "non-functional" pool, was only 1.5GB large, vs 50-150GB on the other effected pools, the tiny pool was still being used, and just because that pool was blovking requests, the whole cluster was unresponsive.So , what do you mean by "non-functional" pool ? how a pool can become non-functional ? and what asures me that tomorrow (just becaue i deleted the 1.5GB pool to fix the whole problem) another pool doesnt becomes non-functional ?Well, you said there were a bunch of slow requests. That can happen any number of ways, if you're overloading the OSDs or something.When there are slow requests, those ops take up OSD memory and throttle, and so they don't let in new messages until the old ones are serviced. This can cascade across a cluster -- because everything is interconnected, clients and OSDs end up with all their requests targeted at the slow OSDs which aren't letting in new IO quickly enough. It's one of the weaknesses of the standard deployment patterns, but it usually doesn't come up unless something else has gone pretty wrong first.As for what actually went wrong here, you haven't provided near enough information and probably can't now that the pool has been deleted. *shrug*-GregCeph Bug ?Another Bug ?Something than can be avoided ?On Tue, Mar 7, 2017 at 2:11 PM, Gregory Farnum <gfarnum@xxxxxxxxxx> wrote:Some facts:The OSDs use a lot of gossip protocols to distribute information.The OSDs limit how many client messages they let in to the system at a time.The OSDs do not distinguish between client ops for different pools (the blocking happens before they have any idea what the target is).So, yes: if you have a non-functional pool and clients keep trying to access it, those requests can fill up the OSD memory queues and block access to other pools as it cascades across the system.On Sun, Mar 5, 2017 at 6:22 PM Alejandro Comisario <alejandro@xxxxxxxxxxx> wrote:______________________________Hi, we have a 7 nodes ubuntu ceph hammer pool (78 OSD to be exact).
This weekend we'be experienced a huge outage from our customers vms
(located on pool CUSTOMERS, replica size 3 ) when lots of OSD's
started to slow request/block PG's on pool PRIVATE ( replica size 1 )
basically all PG's blocked where just one OSD in the acting set, but
all customers on the other pool got their vms almost freezed.
while trying to do basic troubleshooting like doing noout and then
bringing down the OSD that slowed/blocked the most, inmediatelly
another OSD slowed/locked iops on pgs from the same PRIVATE pool, so
we rolled back that change and started to move data around with the
same logic (reweighting down those OSD) with exactly the same result.
So, me made a decition, we decided to delete the pool where all PGS
where slowed/locked allways despite the osd.
Not even 10 secconds passes after the pool deletion, where not only
there were no more degraded PGs, bit also ALL slow iops dissapeared
for ever, and performance from hundreds of vms came to normal
immediately.
I must say that i was kinda scared to see that happen, bascally
because there was only ONE POOL's PGS always slowed, but performance
hit the another pool, so ... did not the PGS that exists on one pool
are not shared by the other ?
If my assertion is true, why OSD's locking iops from one pool's pg
slowed down all other pgs from other pools ?
again, i just deleted a pool that has almost no traffic, because its
pgs were locked and affected pgs on another pool, and as soon as that
happened, the whole cluster came back to normal (and of course,
HEALTH_OK and no slow transaction whatsoever)
please, someone help me understand the gap where i miss something,
since this , as long as my ceph knowledge is concerned, makes no
sense.
PS: i have found someone that , looks like went through the same here:
https://forum.proxmox.com/threads/ceph-osd-failure-causing- proxmox-node-to-crash.20781/
but i still dont understand what happened.
hoping to get the help from the community.
--
Alejandrito._________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
--
Alejandro Comisario
CTO | NUBELIU
E-mail: alejandro@xxxxxxxxxxxCell: +54 9 11 3770 1857
_
www.nubeliu.com--
Alejandro Comisario
CTO | NUBELIU
E-mail: alejandro@xxxxxxxxxxxCell: +54 9 11 3770 1857
_
www.nubeliu.com
| Alejandro Comisario CTO | NUBELIU E-mail: alejandro@xxxxxxxxxxxCell: +54 9 11 3770 1857 _ www.nubeliu.com |
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
