It's pain, but not... :(
We already used your updated lib in dev env... :(
2015-10-30 10:06 GMT+02:00 Wido den Hollander <wido@xxxxxxxx>:
On 29-10-15 16:38, Voloshanenko Igor wrote:
> Hi Wido and all community.
>
> We catched very idiotic issue on our Cloudstack installation, which
> related to ceph and possible to java-rados lib.
>
I think you ran into this one:
https://issues.apache.org/jira/browse/CLOUDSTACK-8879
Cleaning up RBD snapshots for volumes didn't go well and caused the JVM
to crash.
Wido
> cephmon.anolim.net:6789 <http://cephmon.anolim.net:6789>
> So, we have constantly agent crashed (which cause very big problem for
> us... ).
>
> When agent crashed - it's crash JVM. And no event in logs at all.
> We enabled crush dump, and after crash we see next picture:
>
> #grep -A1 "Problematic frame" < /hs_err_pid30260.log
> Problematic frame:
> C [librbd.so.1.0.0+0x5d681]
>
> # gdb /usr/lib/librbd.so.1.0.0 /var/tmp/cores/jsvc.25526.0.core
> (gdb) bt
> ...
> #7 0x00007f30b9a1fed2 in ceph::log::SubsystemMap::should_gather
> (level=<optimized out>, sub=<optimized out>, this=<optimized out>)
> at ./log/SubsystemMap.h:62
> #8 0x00007f30b9a3b693 in ceph::log::SubsystemMap::should_gather
> (this=<optimized out>, sub=<optimized out>, level=<optimized out>)
> at ./log/SubsystemMap.h:61
> #9 0x00007f30b9d879be in ObjectCacher::flusher_entry
> (this=0x7f2fb4017910) at osdc/ObjectCacher.cc:1527
> #10 0x00007f30b9d9851d in ObjectCacher::FlusherThread::entry
> (this=<optimized out>) at osdc/ObjectCacher.h:374
>
> From ceph code, this part executed when flushing cache object... And we
> don;t understand why. Becasue we have absolutely different race
> condition to reproduce it.
>
> As cloudstack have not good implementation yet of snapshot lifecycle,
> sometime, it's happen, that some volumes already marked as EXPUNGED in
> DB and then cloudstack try to delete bas Volume, before it's try to
> unprotect it.
>
> Sure, unprotecting fail, normal exception returned back (fail because
> snap has childs... )
>
> 2015-10-29 09:02:19,401 DEBUG [kvm.resource.KVMHAMonitor]
> (Thread-1304:null) Executing:
> /usr/share/cloudstack-common/scripts/vm/hypervisor/kvm/kvmheartbeat.sh
> -i 10.44.253.13 -p /var/lib/libvirt/PRIMARY -m
> /mnt/93655746-a9ef-394d-95e9-6e62471dd39f -h 10.44.253.11
> 2015-10-29 09:02:19,412 DEBUG [kvm.resource.KVMHAMonitor]
> (Thread-1304:null) Execution is successful.
> 2015-10-29 09:02:19,554 INFO [kvm.storage.LibvirtStorageAdaptor]
> (agentRequest-Handler-5:null) Unprotecting and Removing RBD snapshots of
> image 6789/71b1e2e9-1985-45ca-9ab6-9e5016b86b7c prior to removing the image
> 2015-10-29 09:02:19,571 DEBUG [kvm.storage.LibvirtStorageAdaptor]
> (agentRequest-Handler-5:null) Succesfully connected to Ceph cluster at
> <http://cs2.anolim.net>"},"i-7-106-VM":{"state":"PowerOn","host":"cs2.anolim.net> 2015-10-29 09:02:19,608 DEBUG [kvm.storage.LibvirtStorageAdaptor]
> (agentRequest-Handler-5:null) Unprotecting snapshot
> cloudstack/71b1e2e9-1985-45ca-9ab6-9e5016b86b7c@cloudstack-base-snap
> 2015-10-29 09:02:19,627 DEBUG [kvm.storage.KVMStorageProcessor]
> (agentRequest-Handler-5:null) Failed to delete volume:
> com.cloud.utils.exception.CloudRuntimeException:
> com.ceph.rbd.RbdException: Failed to unprotect snapshot cloudstack-base-snap
> 2015-10-29 09:02:19,628 DEBUG [cloud.agent.Agent]
> (agentRequest-Handler-5:null) Seq 4-1921583831: { Ans: , MgmtId:
> 161344838950, via: 4, Ver: v1, Flags: 10,
> [{"com.cloud.agent.api.Answer":{"result":false,"details":"com.cloud.utils.exception.CloudRuntimeException:
> com.ceph.rbd.RbdException: Failed to unprotect snapshot
> cloudstack-base-snap","wait":0}}] }
> 2015-10-29 09:02:25,722 DEBUG [cloud.agent.Agent]
> (agentRequest-Handler-2:null) Processing command:
> com.cloud.agent.api.GetHostStatsCommand
> 2015-10-29 09:02:25,722 DEBUG [kvm.resource.LibvirtComputingResource]
> (agentRequest-Handler-2:null) Executing: /bin/bash -c idle=$(top -b -n
> 1| awk -F, '/^[%]*[Cc]pu/{$0=$4; gsub(/[^0-9.,]+/,""); print }'); echo $idle
> 2015-10-29 09:02:26,249 DEBUG [kvm.resource.LibvirtComputingResource]
> (agentRequest-Handler-2:null) Execution is successful.
> 2015-10-29 09:02:26,250 DEBUG [kvm.resource.LibvirtComputingResource]
> (agentRequest-Handler-2:null) Executing: /bin/bash -c
> freeMem=$(free|grep cache:|awk '{print $4}');echo $freeMem
> 2015-10-29 09:02:26,254 DEBUG [kvm.resource.LibvirtComputingResource]
> (agentRequest-Handler-2:null) Execution is successful.
>
> BUT, after 20 minutes - agent crashed... If we remove all childs and
> create conditions for cloudstack to delete volume - all OK - no agent
> crash in 20 minutes...
>
> We can't connect this action - Volume delete with agent crashe... Also
> we don't understand why +- 20 minutes need to last, and only then agent
> crashed...
>
> From logs, before crash - only GetVMStats... And then - agent started...
>
> 2015-10-29 09:21:55,143 DEBUG [cloud.agent.Agent] (UgentTask-5:null)
> Sending ping: Seq 4-1343: { Cmd , MgmtId: -1, via: 4, Ver: v1, Flags:
> 11,
> [{"com.cloud.agent.api.PingRoutingCommand":{"newStates":{},"_hostVmStateReport":{"i-881-1117-VM":{"state":"PowerOn","host":"cs2.anolim.net
> <http://cs2.anolim.net>"},"i-1683-1984-VM":{"state":"PowerOn","host":"cs2.anolim.net
> <http://cs2.anolim.net>"},"i-11-504-VM":{"state":"PowerOn","host":"cs2.anolim.net
> <http://cs2.anolim.net>"},"i-325-616-VM":{"state":"PowerOn","host":"cs2.anolim.net
> <http://cs2.anolim.net>"},"i-10-52-VM":{"state":"PowerOn","host":"cs2.anolim.net
> <http://cs2.anolim.net>"},"i-941-1237-VM":{"state":"PowerOn","host":"cs2.anolim.net
> <http://cs2.anolim.net>"}},"_gatewayAccessible":true,"_vnetAccessible":true,"hostType":"Routing","hostId":4,"wait":0}}]
> }
> 2015-10-29 09:21:55,149 DEBUG [cloud.agent.Agent] (Agent-Handler-3:null)
> Received response: Seq 4-1343: { Ans: , MgmtId: 161344838950, via: 4,
> Ver: v1, Flags: 100010,
> [{"com.cloud.agent.api.PingAnswer":{"_command":{"hostType":"Routing","hostId":4,"wait":0},"result":true,"wait":0}}]
> }
> 2015-10-29 09:22:20,755 DEBUG [cloud.agent.Agent]
> (agentRequest-Handler-5:null) Processing command:
> com.cloud.agent.api.GetVmStatsCommand
> 2015-10-29 09:23:02,630 INFO [cloud.agent.AgentShell] (main:null) Agent
> started
> 2015-10-29 09:23:02,632 INFO [cloud.agent.AgentShell] (main:null)
> Implementation Version is 4.3.2
>
>
> Can you please give us direction to go deeper... As we dont udnerstand
> what happened...
>
> Issue reproducing very easy:
>
> (to speed up delete process we set storage.cleanup.interval=10)
>
> 1. Create Volume
> 2. Attach it to VM
> 3. Create snapshot from this volume by hand on ceph
> 4. Do snap_protect by hand on ceph for snap from previous step
> 5. Create child via snap clone
> 6. Deattach volume from VM and delete it
>
> Our problem - that this happen not in such lab situation, but in real
> life...
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
