Status of jobs in etcd doesn't reflect true state of job

[fbalak]

When calling GET /1.0/:cluster_id:/GlusterCreateVolume it returns job_id. When checking job_id through etcd api /v2/keys/queue/:job_id: there is written "status": "finished", but no gluster volume is created. There should be some error state when job fails.

[sankarshanmukhopadhyay]

Have we addressed this topic in any of the recent builds? This may not be a test blocker but the failure condition needs to be handled.

[r0h4n]

We have added detailed job statuses after Dec 5

@fbalak can you reproduce this bug using latest builds?

[fbalak]

@r0h4n When I call API request with invalid data then job won't finish and status remains new or processing. For example calling CreateVolume without specified Volume.bricks keeps job in processing status but it should be failed.
Tested with tendrl-api-1.2-02_24_2017_01_51_03.noarch.

[shtripat]

Sent Tendrl/commons#172 to fix the issue.
This PR marks the job failed is some atom fails execution. Additionally if mandatory parameters are not passed the job is marked as failed.

[sankarshanmukhopadhyay]

With Tendrl/commons#172 merged, is it available in a build? @shtripat - can you add the package NVR which contains the fix?

[shtripat]

The build tendrl-commons-1.2.1-2.el7.centos.noarch.rpm contains the fix which was done 5 days back.

[sankarshanmukhopadhyay]

I am closing this issue. We can revisit it if required.

[mbukatov]

@fbalak Do you agree that this has been fixed? We would need to verify this on a cluster (when we have working builds again) and state the result here.

Based on description here so far, I don't think that enough evidence for closing this issue has been presented.

[mkudlej]

I'm reopening this issue because we cannot be sure that it works. We cannot import cluster( #94 ) or create it so we cannot verify this.

[mkudlej]

@fbalak Please open this issue.

[fbalak]

@mkudlej I am not authorized to do that.
CC: @sankarshanmukhopadhyay

[r0h4n]

This is fixed in v1.2.2

Please verify again

[fbalak]

It seems ok. Failed jobs have 'failed' status. --> closing
Tested with:
tendrl-api-1.2.1-04_03_2017_01_51_03.noarch
tendrl-performance-monitoring-1.2.1-04_03_2017_13_14_13.noarch
tendrl-commons-1.2.1-04_01_2017_00_59_17.noarch