oVirt 4 master storage down - unable to activate after power loss
Dear oVirt users I've been having several problems with my oVirt nodes utilising GlusterFS after simulating a power loss. Our testing setup for oVirt consists of 3 nodes, 2 with GlusterFS storage and 1 just for computing. Everything seems to be setup correct and was working correctly. Then after simulating a power loss the master storage goes down, and therefore the VMs and data center go down as well. I checked the GlusterFS configuration and it seems to be correct (see attachment). I checked the oVirt configuration and it seems to be correct (see attachment). I tried putting several Nodes in Maintenace several times, even putting in maintenance and reinstalling them. Only when putting the nodes in maintenance (and choosing to stop the GlusterFS service) triggers Contending on the other Nodes. Only then there is a chance that a Node goes from Contending to SPM, which does not always happen. After trying this several times I got a main Node to become SPM, but the master storage remains down. When I select the master storage, go in the Data center, select the data center and click Activate, then both the master storage and the data center go in the state Locked for a few seconds and then Inactive again. Then I upgraded to oVirt 4 (oVirt Engine Version: 4.0.0.6-1.el7.centos) and tried everything again, resulting in the same result. I searched online and found this mailing list which is very similar: http://lists.ovirt.org/pipermail/users/2014-August/026905.html . Unfortunately the solution was not posted/mailed. I also found this: https://www.mail-archive.com/users@ovirt.org/msg08105.html suggesting the kernel version is not correct, but I am unsure how to check this. In the attachment I added relevant logs: - GlusterFS service on both main nodes - /var/log/vdsm/vdsm.log , /var/log/vdsm/supervdsm.log , /var/log/messages , /var/log/sanlock.log , /var/log/ovirt-engine/engine.log-20160725.gz , /var/log/ovirt-engine/engine.log of the oVirt engine as attachments (I clicked on Activate on the data center at 9:45:30 am GMT). It can be possible that I need to send different engine logs, please tell me if necessary. Any help would be highly appreciated. Kind regards Robin Vanderveken -- Robin Vanderveken DLT Technologies Tel. +352 691 412 922 -- *Disclaimer* *Les informations contenues dans ce message sont destinées exclusivement à l’usage de la personne à laquelle elles sont adressées. Elles peuvent contenir des données confidentielles et/ou privilégiées et/ou protégées par des droits de propriété intellectuelle ou d’autres lois. Si vous avez reçu ce message par erreur, veuillez le détruire, ainsi que toutes ses annexes, et notifier son expéditeur ; il vous est interdit de copier ou d’utiliser ce messages ou ses annexes ou de divulguer son contenu à quiconque. La transmission de données par e-mail ne peut être garantie comme un moyen sûr et infaillible, ni comme étant exempt de tout virus. L’expéditeur décline toute responsabilité en cas de perte ou de dommage résultant de ce message ou de son utilisation, quelle qu’elle soit. * *The information contained in this message is intended for the addressee only and may contain confidential and/or privileged information and/or information protected by intellectual property rights or other legal rules. If you are not the intended recipient, please delete this message and any attachment to it and notify the sender; you may not copy or use this message or its attachments in any way nor disclose its contents to anyone. Emails cannot be guaranteed to be secure or to be error or virus free. No liability is accepted by the sender for any loss damage arising in any way from this message or its use. *
First off, a replica 2 gluster volume is not recommended if you want to survive power loss - you need to have a replica 3 gluster volume.
From engine logs, I see "Could not connect host to Data Center(Storage issue)" but there are no errors in the attached vdsm.log. Can you provide the relevant vdsm log and also the gluster mount logs?
What's the output of "gluster volume heal <volname> info" ? ----- Original Message -----
From: "Robin Vanderveken" <rvanderveken@dlttechnologies.com> To: users@ovirt.org Sent: Monday, July 25, 2016 2:01:24 PM Subject: [ovirt-users] oVirt 4 master storage down - unable to activate after power loss
Dear oVirt users
I've been having several problems with my oVirt nodes utilising GlusterFS after simulating a power loss.
Our testing setup for oVirt consists of 3 nodes, 2 with GlusterFS storage and 1 just for computing. Everything seems to be setup correct and was working correctly. Then after simulating a power loss the master storage goes down, and therefore the VMs and data center go down as well.
I checked the GlusterFS configuration and it seems to be correct (see attachment). I checked the oVirt configuration and it seems to be correct (see attachment). I tried putting several Nodes in Maintenace several times, even putting in maintenance and reinstalling them. Only when putting the nodes in maintenance (and choosing to stop the GlusterFS service) triggers Contending on the other Nodes. Only then there is a chance that a Node goes from Contending to SPM, which does not always happen. After trying this several times I got a main Node to become SPM, but the master storage remains down. When I select the master storage, go in the Data center, select the data center and click Activate, then both the master storage and the data center go in the state Locked for a few seconds and then Inactive again.
Then I upgraded to oVirt 4 ( oVirt Engine Version: 4.0.0.6-1.el7.centos) and tried everything again, resulting in the same result.
I searched online and found this mailing list which is very similar: http://lists.ovirt.org/pipermail/users/2014-August/026905.html . Unfortunately the solution was not posted/mailed. I also found this: https://www.mail-archive.com/users@ovirt.org/msg08105.html suggesting the kernel version is not correct, but I am unsure how to check this.
In the attachment I added relevant logs: - GlusterFS service on both main nodes - /var/log/vdsm/vdsm.log , /var/log/vdsm/supervdsm.log , /var/log/messages , /var/log/sanlock.log , /var/log/ovirt-engine/engine. log-20160725.gz , /var/log/ovirt-engine/engine.log of the oVirt engine as attachments (I clicked on Activate on the data center at 9:45:30 am GMT). It can be possible that I need to send different engine logs, please tell me if necessary. Any help would be highly appreciated.
Kind regards Robin Vanderveken
-- Robin Vanderveken DLT Technologies Tel. +352 691 412 922
Disclaimer
Les informations contenues dans ce message sont destinées exclusivement à l’usage de la personne à laquelle elles sont adressées. Elles peuvent contenir des données confidentielles et/ou privilégiées et/ou protégées par des droits de propriété intellectuelle ou d’autres lois. Si vous avez reçu ce message par erreur, veuillez le détruire, ainsi que toutes ses annexes, et notifier son expéditeur ; il vous est interdit de copier ou d’utiliser ce messages ou ses annexes ou de divulguer son contenu à quiconque. La transmission de données par e-mail ne peut être garantie comme un moyen sûr et infaillible, ni comme étant exempt de tout virus. L’expéditeur décline toute responsabilité en cas de perte ou de dommage résultant de ce message ou de son utilisation, quelle qu’elle soit.
The information contained in this message is intended for the addressee only and may contain confidential and/or privileged information and/or information protected by intellectual property rights or other legal rules. If you are not the intended recipient, please delete this message and any attachment to it and notify the sender; you may not copy or use this message or its attachments in any way nor disclose its contents to anyone. Emails cannot be guaranteed to be secure or to be error or virus free. No liability is accepted by the sender for any loss damage arising in any way from this message or its use.
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Your command was very helpful! It seems to be a split-brain error! I should be able to fix this, I'll let you know when it's fixed. We are aware that 2 gluster volumes is not recommended, we didn't know a minimum of 3 is needed to counter a power loss. The output of "gluster volume heal <volume_name> info" is: [root@ovirt-node2 ~]# gluster volume heal vol1 info Brick 10.0.0.194:/gluster-bricks/vol1/vol1 /9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/ids - Is in split-brain /__DIRECT_IO_TEST__ - Is in split-brain /9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/inbox - Is in split-brain /9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/leases - Is in split-brain Status: Connected Number of entries: 4 Brick 10.0.0.199:/gluster-bricks/vol1/vol1 <gfid:b9d48873-846d-4964-93c7-0a662de04d22> - Is in split-brain /__DIRECT_IO_TEST__ - Is in split-brain <gfid:cbc7b097-e600-43d3-8f20-9626f6d0239b> - Is in split-brain <gfid:cfa177c0-91fa-4649-b99b-73e03ddbc756> - Is in split-brain Status: Connected Number of entries: 4 I also added GlusterFS logs, I was unsure which you needed so I added as many as possible. Kind regards Robin Vanderveken On 25 July 2016 at 15:49, Sahina Bose <sabose@redhat.com> wrote:
First off, a replica 2 gluster volume is not recommended if you want to survive power loss - you need to have a replica 3 gluster volume.
From engine logs, I see "Could not connect host to Data Center(Storage issue)" but there are no errors in the attached vdsm.log. Can you provide the relevant vdsm log and also the gluster mount logs?
What's the output of "gluster volume heal <volname> info" ?
From: "Robin Vanderveken" <rvanderveken@dlttechnologies.com> To: users@ovirt.org Sent: Monday, July 25, 2016 2:01:24 PM Subject: [ovirt-users] oVirt 4 master storage down - unable to activate after power loss
Dear oVirt users
I've been having several problems with my oVirt nodes utilising GlusterFS after simulating a power loss.
Our testing setup for oVirt consists of 3 nodes, 2 with GlusterFS storage and 1 just for computing. Everything seems to be setup correct and was working correctly. Then after simulating a power loss the master storage goes down, and therefore the VMs and data center go down as well.
I checked the GlusterFS configuration and it seems to be correct (see attachment). I checked the oVirt configuration and it seems to be correct (see attachment). I tried putting several Nodes in Maintenace several times, even putting in maintenance and reinstalling them. Only when putting the nodes in maintenance (and choosing to stop the GlusterFS service)
Contending on the other Nodes. Only then there is a chance that a Node goes from Contending to SPM, which does not always happen. After trying this several times I got a main Node to become SPM, but the master storage remains down. When I select the master storage, go in the Data center, select the data center and click Activate, then both the master storage and the data center go in the state Locked for a few seconds and then Inactive again.
Then I upgraded to oVirt 4 ( oVirt Engine Version: 4.0.0.6-1.el7.centos) and tried everything again, resulting in the same result.
I searched online and found this mailing list which is very similar: http://lists.ovirt.org/pipermail/users/2014-August/026905.html . Unfortunately the solution was not posted/mailed. I also found this: https://www.mail-archive.com/users@ovirt.org/msg08105.html suggesting
kernel version is not correct, but I am unsure how to check this.
In the attachment I added relevant logs: - GlusterFS service on both main nodes - /var/log/vdsm/vdsm.log , /var/log/vdsm/supervdsm.log , /var/log/messages , /var/log/sanlock.log , /var/log/ovirt-engine/engine. log-20160725.gz , /var/log/ovirt-engine/engine.log of the oVirt engine as attachments (I clicked on Activate on the data center at 9:45:30 am GMT). It can be possible that I need to send different engine logs, please tell me if necessary. Any help would be highly appreciated.
Kind regards Robin Vanderveken
-- Robin Vanderveken DLT Technologies Tel. +352 691 412 922
Disclaimer
Les informations contenues dans ce message sont destinées exclusivement à l’usage de la personne à laquelle elles sont adressées. Elles peuvent contenir des données confidentielles et/ou privilégiées et/ou protégées
des droits de propriété intellectuelle ou d’autres lois. Si vous avez reçu ce message par erreur, veuillez le détruire, ainsi que toutes ses annexes, et notifier son expéditeur ; il vous est interdit de copier ou d’utiliser ce messages ou ses annexes ou de divulguer son contenu à quiconque. La transmission de données par e-mail ne peut être garantie comme un moyen sûr et infaillible, ni comme étant exempt de tout virus. L’expéditeur décline toute responsabilité en cas de perte ou de dommage résultant de ce message ou de son utilisation, quelle qu’elle soit.
The information contained in this message is intended for the addressee only and may contain confidential and/or privileged information and/or information protected by intellectual property rights or other legal rules. If you are not the intended recipient, please delete this message and any attachment to it and notify the sender; you may not copy or use this message or its attachments in any way nor disclose its contents to anyone. Emails cannot be guaranteed to be secure or to be error or virus free. No
is accepted by the sender for any loss damage arising in any way from
----- Original Message ----- triggers the par liability this
message or its use.
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
-- Robin Vanderveken DLT Technologies Tel. +352 691 412 922 -- *Disclaimer* *Les informations contenues dans ce message sont destinées exclusivement à l’usage de la personne à laquelle elles sont adressées. Elles peuvent contenir des données confidentielles et/ou privilégiées et/ou protégées par des droits de propriété intellectuelle ou d’autres lois. Si vous avez reçu ce message par erreur, veuillez le détruire, ainsi que toutes ses annexes, et notifier son expéditeur ; il vous est interdit de copier ou d’utiliser ce messages ou ses annexes ou de divulguer son contenu à quiconque. La transmission de données par e-mail ne peut être garantie comme un moyen sûr et infaillible, ni comme étant exempt de tout virus. L’expéditeur décline toute responsabilité en cas de perte ou de dommage résultant de ce message ou de son utilisation, quelle qu’elle soit. * *The information contained in this message is intended for the addressee only and may contain confidential and/or privileged information and/or information protected by intellectual property rights or other legal rules. If you are not the intended recipient, please delete this message and any attachment to it and notify the sender; you may not copy or use this message or its attachments in any way nor disclose its contents to anyone. Emails cannot be guaranteed to be secure or to be error or virus free. No liability is accepted by the sender for any loss damage arising in any way from this message or its use. *
I resolved the issue. First I put the nodes in Maintenance mode (not stopping Gluster service). Then because the files that were in split-brain mode were not important I set the hex value to 0 on both nodes. I guess I could have used something like "getfattr -d -m . -e hex /gluster-bricks/vol1/vol1/9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/ids" and find the correct hex and such, the following article explains it nicely: https://access.redhat.com/documentation/en-US/Red_Hat_Storage/3/html/Adminis... . [root@oVirt-Node1 ~]# setfattr -n trusted.afr.vol1-client-1 -v 0x000000000000000000000000 /gluster-bricks/vol1/vol1/9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/ids [root@oVirt-Node1 ~]# setfattr -n trusted.afr.vol1-client-1 -v 0x000000000000000000000000 /gluster-bricks/vol1/vol1 / __DIRECT_IO_TEST__ [root@oVirt-Node1 ~]# setfattr -n trusted.afr.vol1-client-1 -v 0x000000000000000000000000 /gluster-bricks/vol1/vol1/9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/inbox [root@oVirt-Node1 ~]# setfattr -n trusted.afr.vol1-client-1 -v 0x000000000000000000000000 /gluster-bricks/vol1/vol1/9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/leases [root@ovirt-node2 ~]# setfattr -n trusted.afr.vol1-client-0 -v 0x000000000000000000000000 /gluster-bricks/vol1/vol1/9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/ids [root@ovirt-node2 ~]# setfattr -n trusted.afr.vol1-client-0 -v 0x000000000000000000000000 /gluster-bricks/vol1/vol1 / __DIRECT_IO_TEST__ [root@ovirt-node2 ~]# setfattr -n trusted.afr.vol1-client-0 -v 0x000000000000000000000000 /gluster-bricks/vol1/vol1/9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/inbox [root@ovirt-node2 ~]# setfattr -n trusted.afr.vol1-client-0 -v 0x000000000000000000000000 /gluster-bricks/vol1/vol1/9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/leases The heal command now gives no errors. [root@oVirt-Node1 ~]# gluster volume heal vol1 info Brick 10.0.0.194:/gluster-bricks/vol1/vol1 /__DIRECT_IO_TEST__ /__DIRECT_IO_TEST__ Status: Connected Number of entries: 2 Brick 10.0.0.199:/gluster-bricks/vol1/vol1 /__DIRECT_IO_TEST__ /__DIRECT_IO_TEST__ Status: Connected Number of entries: 2 I activated the nodes again in oVirt and activated the data center (going in Storage > master volume > Data Center > Activate). Now it goes to Locked and to Up. Now the SPM is automatically selected in the nodes. Thank you very much!!! Kind regards Robin Vanderveken On 25 July 2016 at 16:10, Robin Vanderveken < rvanderveken@dlttechnologies.com> wrote:
Your command was very helpful! It seems to be a split-brain error! I should be able to fix this, I'll let you know when it's fixed.
We are aware that 2 gluster volumes is not recommended, we didn't know a minimum of 3 is needed to counter a power loss.
The output of "gluster volume heal <volume_name> info" is:
[root@ovirt-node2 ~]# gluster volume heal vol1 info Brick 10.0.0.194:/gluster-bricks/vol1/vol1 /9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/ids - Is in split-brain
/__DIRECT_IO_TEST__ - Is in split-brain
/9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/inbox - Is in split-brain
/9d1a204c-1b1a-450f-986d-a8c84babb4c1/dom_md/leases - Is in split-brain
Status: Connected Number of entries: 4
Brick 10.0.0.199:/gluster-bricks/vol1/vol1 <gfid:b9d48873-846d-4964-93c7-0a662de04d22> - Is in split-brain
/__DIRECT_IO_TEST__ - Is in split-brain
<gfid:cbc7b097-e600-43d3-8f20-9626f6d0239b> - Is in split-brain
<gfid:cfa177c0-91fa-4649-b99b-73e03ddbc756> - Is in split-brain
Status: Connected Number of entries: 4
I also added GlusterFS logs, I was unsure which you needed so I added as many as possible.
Kind regards Robin Vanderveken
On 25 July 2016 at 15:49, Sahina Bose <sabose@redhat.com> wrote:
First off, a replica 2 gluster volume is not recommended if you want to survive power loss - you need to have a replica 3 gluster volume.
From engine logs, I see "Could not connect host to Data Center(Storage issue)" but there are no errors in the attached vdsm.log. Can you provide the relevant vdsm log and also the gluster mount logs?
What's the output of "gluster volume heal <volname> info" ?
From: "Robin Vanderveken" <rvanderveken@dlttechnologies.com> To: users@ovirt.org Sent: Monday, July 25, 2016 2:01:24 PM Subject: [ovirt-users] oVirt 4 master storage down - unable to activate after power loss
Dear oVirt users
I've been having several problems with my oVirt nodes utilising GlusterFS after simulating a power loss.
Our testing setup for oVirt consists of 3 nodes, 2 with GlusterFS storage and 1 just for computing. Everything seems to be setup correct and was working correctly. Then after simulating a power loss the master storage goes down, and therefore the VMs and data center go down as well.
I checked the GlusterFS configuration and it seems to be correct (see attachment). I checked the oVirt configuration and it seems to be correct (see attachment). I tried putting several Nodes in Maintenace several times, even putting in maintenance and reinstalling them. Only when putting the nodes in maintenance (and choosing to stop the GlusterFS service)
Contending on the other Nodes. Only then there is a chance that a Node goes from Contending to SPM, which does not always happen. After trying this several times I got a main Node to become SPM, but the master storage remains down. When I select the master storage, go in the Data center, select the data center and click Activate, then both the master storage and the data center go in the state Locked for a few seconds and then Inactive again.
Then I upgraded to oVirt 4 ( oVirt Engine Version: 4.0.0.6-1.el7.centos) and tried everything again, resulting in the same result.
I searched online and found this mailing list which is very similar: http://lists.ovirt.org/pipermail/users/2014-August/026905.html . Unfortunately the solution was not posted/mailed. I also found this: https://www.mail-archive.com/users@ovirt.org/msg08105.html suggesting
kernel version is not correct, but I am unsure how to check this.
In the attachment I added relevant logs: - GlusterFS service on both main nodes - /var/log/vdsm/vdsm.log , /var/log/vdsm/supervdsm.log , /var/log/messages , /var/log/sanlock.log , /var/log/ovirt-engine/engine. log-20160725.gz , /var/log/ovirt-engine/engine.log of the oVirt engine as attachments (I clicked on Activate on the data center at 9:45:30 am GMT). It can be possible that I need to send different engine logs, please tell me if necessary. Any help would be highly appreciated.
Kind regards Robin Vanderveken
-- Robin Vanderveken DLT Technologies Tel. +352 691 412 922
Disclaimer
Les informations contenues dans ce message sont destinées exclusivement à l’usage de la personne à laquelle elles sont adressées. Elles peuvent contenir des données confidentielles et/ou privilégiées et/ou protégées
des droits de propriété intellectuelle ou d’autres lois. Si vous avez reçu ce message par erreur, veuillez le détruire, ainsi que toutes ses annexes, et notifier son expéditeur ; il vous est interdit de copier ou d’utiliser ce messages ou ses annexes ou de divulguer son contenu à quiconque. La transmission de données par e-mail ne peut être garantie comme un moyen sûr et infaillible, ni comme étant exempt de tout virus. L’expéditeur décline toute responsabilité en cas de perte ou de dommage résultant de ce message ou de son utilisation, quelle qu’elle soit.
The information contained in this message is intended for the addressee only and may contain confidential and/or privileged information and/or information protected by intellectual property rights or other legal rules. If you are not the intended recipient, please delete this message and any attachment to it and notify the sender; you may not copy or use this message or its attachments in any way nor disclose its contents to anyone. Emails cannot be guaranteed to be secure or to be error or virus free. No
is accepted by the sender for any loss damage arising in any way from
----- Original Message ----- triggers the par liability this
message or its use.
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
-- Robin Vanderveken DLT Technologies Tel. +352 691 412 922
-- Robin Vanderveken DLT Technologies Tel. +352 691 412 922 -- *Disclaimer* *Les informations contenues dans ce message sont destinées exclusivement à l’usage de la personne à laquelle elles sont adressées. Elles peuvent contenir des données confidentielles et/ou privilégiées et/ou protégées par des droits de propriété intellectuelle ou d’autres lois. Si vous avez reçu ce message par erreur, veuillez le détruire, ainsi que toutes ses annexes, et notifier son expéditeur ; il vous est interdit de copier ou d’utiliser ce messages ou ses annexes ou de divulguer son contenu à quiconque. La transmission de données par e-mail ne peut être garantie comme un moyen sûr et infaillible, ni comme étant exempt de tout virus. L’expéditeur décline toute responsabilité en cas de perte ou de dommage résultant de ce message ou de son utilisation, quelle qu’elle soit. * *The information contained in this message is intended for the addressee only and may contain confidential and/or privileged information and/or information protected by intellectual property rights or other legal rules. If you are not the intended recipient, please delete this message and any attachment to it and notify the sender; you may not copy or use this message or its attachments in any way nor disclose its contents to anyone. Emails cannot be guaranteed to be secure or to be error or virus free. No liability is accepted by the sender for any loss damage arising in any way from this message or its use. *
participants (2)
-
Robin Vanderveken -
Sahina Bose