Re: [one-users] Delete Recreate fails in case of Host-Error
This is now fixed for the next release. The snapshot will be reused, the only potential problem is that reusing the same volume with old images from a previous installation, may lead to inconsistent copies. This is an anomalous situation, and reusing the snapshot is probably faster and safer. Cheers Ruben On Sat Dec 13 2014 at 8:58:41 AM Damon (Albino Geek) albinog...@gmail.com wrote: On Fri, 12 Dec 2014 04:15:34 -0800, Fabian Zimmermann dev@gmail.com wrote: Hi, Ah! Thanks I just misread the oned.conf Nevertheless, I used -r and assumed it would re-create the VM, but this failed if you use (ceph) shared storage, because clone will abort if previous cleanup failed, so there is (in my opinion) a bug, because tm should handle this by removing or using the old snap/disk, isn't it? Fabian I do think that it should be better handled by the CephFS TM. ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] Delete Recreate fails in case of Host-Error
Hi Fabian, In OpenNebula 4.10 if the VM is in UNKNOWN it should go directly to boot (bypassing CLEANUP and PROLOG) , provided you are using shared storage... Cheers Ruben On Fri Dec 12 2014 at 4:16:11 AM Fabian Zimmermann dev@gmail.com wrote: Hi, I just setup our fencing system and everything is working so far, but if I use the HOST-ERROR-Hook to deleterecreate the VMs this will fail. The CLEANUP is tried on the broken host - of course it will fail - and the followed CLONE will fail, because of already existing snapshots/disks. I just workaround this - see attached patch, which will just skip clone if disk/snap already exists, but what's the correct way to handle this? Thanks a lot, Fabian Zimmermann ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] Delete Recreate fails in case of Host-Error
Totally, tm-ceph should work in this case... The issue for this: http://dev.opennebula.org/issues/3446 THANKS your feedback!!! Ruben On Fri Dec 12 2014 at 1:15:38 PM Fabian Zimmermann dev@gmail.com wrote: Hi, Am 12.12.14 13:04, schrieb Ruben S. Montero: If you have shared storage, you can simply issue onevm resched (I think the hook can do that with the -m option). If there is no shared storage then you have to go through the delete-recreate step, that will fail for the host as it is down. The host should be cleaned up manually once it is online again. Ah! Thanks I just misread the oned.conf Nevertheless, I used -r and assumed it would re-create the VM, but this failed if you use (ceph) shared storage, because clone will abort if previous cleanup failed, so there is (in my opinion) a bug, because tm should handle this by removing or using the old snap/disk, isn't it? Fabian ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] Delete Recreate fails in case of Host-Error
Hi Ruben, Am 12.12.14 11:20, schrieb Ruben S. Montero: In OpenNebula 4.10 if the VM is in UNKNOWN it should go directly to boot (bypassing CLEANUP and PROLOG) , provided you are using shared storage... we are using 4.10.1 and it looks like CLEANUP is executed isn't it? -- Wed Dec 10 16:38:45 2014 [Z0][LCM][I]: New VM state is RUNNING Thu Dec 11 09:16:01 2014 [Z0][LCM][I]: New VM state is UNKNOWN Thu Dec 11 09:23:32 2014 [Z0][LCM][I]: New VM state is CLEANUP. Thu Dec 11 09:23:32 2014 [Z0][DiM][I]: New VM state is PENDING -- As written, i'm just using the ERR-Host-Hook to execute the host-error.rb and deleterecreate the VMs. Fabian ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] Delete Recreate fails in case of Host-Error
Hi Fabian, If you have shared storage, you can simply issue onevm resched (I think the hook can do that with the -m option). If there is no shared storage then you have to go through the delete-recreate step, that will fail for the host as it is down. The host should be cleaned up manually once it is online again. Cheers On Fri Dec 12 2014 at 12:44:50 PM Fabian Zimmermann dev@gmail.com wrote: Hi Ruben, Am 12.12.14 11:20, schrieb Ruben S. Montero: In OpenNebula 4.10 if the VM is in UNKNOWN it should go directly to boot (bypassing CLEANUP and PROLOG) , provided you are using shared storage... we are using 4.10.1 and it looks like CLEANUP is executed isn't it? -- Wed Dec 10 16:38:45 2014 [Z0][LCM][I]: New VM state is RUNNING Thu Dec 11 09:16:01 2014 [Z0][LCM][I]: New VM state is UNKNOWN Thu Dec 11 09:23:32 2014 [Z0][LCM][I]: New VM state is CLEANUP. Thu Dec 11 09:23:32 2014 [Z0][DiM][I]: New VM state is PENDING -- As written, i'm just using the ERR-Host-Hook to execute the host-error.rb and deleterecreate the VMs. Fabian ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] Delete Recreate fails in case of Host-Error
On Fri, 12 Dec 2014 04:15:34 -0800, Fabian Zimmermann dev@gmail.com wrote: Hi, Ah! Thanks I just misread the oned.conf Nevertheless, I used -r and assumed it would re-create the VM, but this failed if you use (ceph) shared storage, because clone will abort if previous cleanup failed, so there is (in my opinion) a bug, because tm should handle this by removing or using the old snap/disk, isn't it? Fabian I do think that it should be better handled by the CephFS TM. ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
[one-users] Delete Recreate fails in case of Host-Error
Hi, I just setup our fencing system and everything is working so far, but if I use the HOST-ERROR-Hook to deleterecreate the VMs this will fail. The CLEANUP is tried on the broken host - of course it will fail - and the followed CLONE will fail, because of already existing snapshots/disks. I just workaround this - see attached patch, which will just skip clone if disk/snap already exists, but what's the correct way to handle this? Thanks a lot, Fabian Zimmermann --- clone.old 2014-12-11 13:13:52.001648056 +0100 +++ clone 2014-12-11 09:32:18.882111020 +0100 @@ -76,14 +76,15 @@ CLONE_CMD=$(cat EOF set -e -RBD_FORMAT=\$($RBD info $SRC_PATH | sed -n 's/.*format: // p') - -if [ \$RBD_FORMAT = 2 ]; then -$RBD snap create $SRC_PATH@$RBD_SNAP -$RBD snap protect $SRC_PATH@$RBD_SNAP -$RBD clone $SRC_PATH@$RBD_SNAP $RBD_DST -else -$RBD copy $SRC_PATH $RBD_DST +if ! $RBD info $RBD_DST; then +RBD_FORMAT=\$($RBD info $SRC_PATH | sed -n 's/.*format: // p') +if [ \$RBD_FORMAT = 2 ]; then +$RBD snap create $SRC_PATH@$RBD_SNAP +$RBD snap protect $SRC_PATH@$RBD_SNAP +$RBD clone $SRC_PATH@$RBD_SNAP $RBD_DST +else +$RBD copy $SRC_PATH $RBD_DST +fi fi EOF ) ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org