CA-379329: check for missing iSCSI sessions and reconnect #627
Conversation
drivers/sr_health_check.py
Outdated
| import util | ||
|
|
||
|
|
||
| def check_iscsi_sessions(): |
There was a problem hiding this comment.
This feels a bit weird to me. The function is about checking iSCSI sessions but then it calls a generic check_sr() method. Would it be better to just call check_sr() on every SR and then let the SR decide what to check (if anything)?
|
I've just realised that we already have |
drivers/sr_health_check.py
Outdated
|
|
||
| sr_uuid = srs[sr]['uuid'] | ||
| sr_obj = SR.SR.from_uuid(session, sr_uuid) | ||
| sr_obj.check_sr(sr_uuid) |
There was a problem hiding this comment.
What if you have a SMAPIv3 SR here, will that work, or do you need to filter them out?
There was a problem hiding this comment.
You can find a list of SMAPIv1 SRs with something like this (can be translated into python API calls instead of xe):
xe sm-list required-api-version=1.0 params=type --minimal|tr ',' '\n'|xargs -n1 -I% xe sr-list type=% --minimal|tr ',' '\n'|xargs echo
There was a problem hiding this comment.
That's a very good point, originally it only looked for lvmoiscsi SRs and then after addressing #627 (comment) it exposes that problem.
There was a problem hiding this comment.
Now fixed and checked with concurrent LVMoISCSI and GFS2 (on iSCSI) SRs
If, when an iSCSI SR is attached, some of the paths are not reachable they will not be subsequently connected should the connectivity issues be resolved, without detaching the SR or rebooting the host. Neither of these options give a good experience for users of the software. Add a systemd timer which will, periodically (default 10 minutes), ask each SR to do SR specific "health" checks. Currently only a checker for the iSCSI sessions is present but more can be added to this pattern in the future as needs arise. Signed-off-by: Mark Syms <[email protected]>
If, when an iSCSI SR is attached, some of the paths are not reachable they will not be subsequently connected should the connectivity issues be resolved, without detaching the SR or rebooting the host. Neither of these options give a good experience for users of the software. Add a systemd timer which will, periodically (default 10 minutes), ask each SR to do SR specific "health" checks. Currently only a checker for the iSCSI sessions is present but more can be added to this pattern in the future as needs arise.