TOPIC:

Hi,

After the last update and reboot, the SCSI ID is missing and the storage remains disconnected. The problem can be solved by disabling the autostart of the "iscsi-ha" service and manually start it after some time, as recommended in the topic www.halizard.com/forum/software-support/...-unplugged-on-reboot


[06:26 XCP3 ~]# ls /dev/disk/by-scsid/
35000cca095c0f118 35000cca095d37cec 35000cca25ccc4cb1 350025388a0829f14 35002538c409e66f9

[06:29 XCP3 ~]# xe pbd-plug uuid=881b327f-bfae-a91a-f39c-40ded0bcd5a0
Error code: SR_BACKEND_FAILURE_47
Error parameters: , The SR is not available [opterr=[Errno 2] No such file or directory: '/dev/disk/by-scsid/360000000000000000e000000000100d2'],


update / yum.log
Jan 08 05:55:15 Updated: xen-libs-4.13.5-9.38.1.xcpng8.2.x86_64
Jan 08 05:55:15 Updated: xcp-ng-release-presets-8.2.1-10.x86_64
Jan 08 05:55:17 Installed: mpi3mr-module-8.6.1.0.0-1.xcpng8.2.x86_64
Jan 08 05:55:17 Installed: igc-module-5.10.200-1.xcpng8.2.x86_64
Jan 08 05:55:18 Updated: xen-hypervisor-4.13.5-9.38.1.xcpng8.2.x86_64
Jan 08 05:55:18 Updated: xen-dom0-libs-4.13.5-9.38.1.xcpng8.2.x86_64
Jan 08 05:55:18 Updated: xen-tools-4.13.5-9.38.1.xcpng8.2.x86_64
Jan 08 05:55:18 Installed: r8125-module-9.012.03-1.xcpng8.2.x86_64
Jan 08 05:55:18 Updated: vendor-drivers-1.0.2-1.6.xcpng8.2.x86_64
Jan 08 05:55:19 Updated: xen-dom0-tools-4.13.5-9.38.1.xcpng8.2.x86_64
Jan 08 05:55:19 Updated: gpumon-0.18.0-11.2.xcpng8.2.x86_64
Jan 08 05:55:22 Updated: xcp-ng-release-config-8.2.1-10.x86_64
Jan 08 05:55:22 Updated: xcp-ng-release-8.2.1-10.x86_64
Jan 08 05:55:26 Updated: linux-firmware-20190314-10.2.xcpng8.2.noarch
Jan 08 05:55:27 Updated: tzdata-2023c-1.el7.noarch

Thanks Michal,
We are looking into this race condition. Currently, the iscsi-ha daemon waits for xapi to place a token on the file system to indicate that xapi has finished starting. This is clearly creating a race condition causing iscsi-ha to start prematurely.

Thanks for the answer. If you need some more tests or logs, I can provide it

Hi Salvatore,

I tried another server reboot yesterday and the original solution (disabling iscsi-ha autostart) is now not working. But I tried other things and I found that stopping iscsid (when iscsi-ha, watchdog, tgtd are stopped) and after about 30sec starting it will help and the ids of the disks in /dev/disk/by-scsid/ will appear.
Wouldn't updating iscsid (e.g. iscsi-initiator-utils package) help?

On the same situation here... just upgraded the server and got this errors.. trying just now to fix it.

Salvatore.. if are online, do you want to make a call to get a look ??

Note: Just in case... only disabling the services and rebooting wasnt enough.. we had to follow the workaround from Adamec Michal (stop the iscisd and wait and start all manually) to have it working.

Hi, we're also affected by this error since
the latest "yum update" xcp-package updates.

Error code: SR_BACKEND_FAILURE_47
Error parameters: , The SR is not available [opterr=[Errno 2] No such file or directory: '/dev/disk/by-scsid/360000000000000000e000000000100d2'],


disabling iscsi-ha and iscsi-ha-watchdog services did NOT work for us.

But when iscsi-ha is started and this error is logged,
we were able to get everything up and runnig by runnig this command on the master node:

systemctl stop tgtd

Then we took a look at this directory: ls -lah /dev/disk/by-scsid/
After about 15 seconds the missing device "/dev/disk/by-scsid/360000000000000000e0000000001000a suddenly appeared.
We also noticed that tgtd was restarted automatically, PDB was plugged and the VM startet autmatically.

So currently our workaround is to run "systemctl stop tgtd" on the master after rebooting xcp hosts.

Kind regards
Robert Schuh