After some analysis, I think the bug we are seeing here is https://bugzilla.redhat.com/show_bug.cgi?id=1588061
This applies for suspend/resume and also for a snapshot with memory.
Following the steps and considering that the iscsi storage domain is only 20GB, this should be the reason for reaching ~4GB free space

On Fri, Nov 30, 2018 at 10:01 PM Raz Tamir <ratamir@redhat.com> wrote:


On Fri, Nov 30, 2018, 21:57 Ryan Barry <rbarry@redhat.com wrote:


On Fri, Nov 30, 2018 at 2:31 PM Raz Tamir <ratamir@redhat.com> wrote:


On Fri, Nov 30, 2018, 19:33 Dafna Ron <dron@redhat.com wrote:
Hi,

This mail is to provide the current status of CQ and allow people to review status before and after the weekend.
Please refer to below colour map for further information on the meaning of the colours.

CQ-4.2: RED (#1)

I checked last date ovirt-engine and vdsm passed and moved packages to tested as they are the bigger projects and it was on the 27-11-218.

We have been having sporadic failures for most of the projects on test check_snapshot_with_memory.
We have deducted that this is caused by a code regression in storage based on the following things:
1.Evgheni and Gal helped debug this issue to rule out lago and infra issue as the cause of failure and both determined the issue is a code regression - most likely in storage.
2. The failure only happens on 4.2 branch.
3. the failure itself is cannot run a vm due to low disk space in storage domain and we cannot see any failures which would leave any leftovers in the storage domain.
Can you please share the link to the execution? 

Here's an example of one run:

The iSCSI storage domain starts emitting warnings about low storage space immediately after removing the VmPool, but it's possible that the storage domain is filling before that from some other call prior to that which is still running, possibly the VM import.
Thanks Ryan, I'll try to help with debugging this issue 
 

Dan and Ryan are actively involved in trying to find the regression but the consensus is that this is a storage related regression and we are having a problem getting the storage team to join us in debugging the issue.

I prepared a patch to skip the test in case we cannot get cooperation from storage team and resolve this regression in the next few days:
https://gerrit.ovirt.org/#/c/95889/

CQ-Master: YELLOW (#1)

We have failures which CQ is still bisecting and until its done we cannot point to any specific failing projects.


Happy week!
Dafna


-------------------------------------------------------------------------------------------------------------------
COLOUR MAP

Green = job has been passing successfully

** green for more than 3 days may suggest we need a review of our test coverage


  1. 1-3 days       GREEN (#1)

  2. 4-7 days       GREEN (#2)

  3. Over 7 days GREEN (#3)


Yellow = intermittent failures for different projects but no lasting or current regressions

** intermittent would be a healthy project as we expect a number of failures during the week

** I will not report any of the solved failures or regressions.


  1. Solved job failures        YELLOW (#1)

  2. Solved regressions      YELLOW (#2)


Red = job has been failing

** Active Failures. The colour will change based on the amount of time the project/s has been broken. Only active regressions would be reported.


  1. 1-3 days      RED (#1)

  2. 4-7 days      RED (#2)

  3. Over 7 days RED (#3)




--

Ryan Barry

Associate Manager - RHV Virt/SLA

rbarry@redhat.com    M: +16518159306     IM: rbarry



--


Raz Tamir
Manager, RHV QE