Jira postitive index count

[Greg Warner (Amazon)]

Hi,

I actively monitor the lucene indexes across our Jira Data Center hosts in order to ensure a consistent and reliable environment for our end users. I do this through an in-house custom add-on.

Right now for Jira:

One of the nodes is reporting positive lucene index difference (there is more in the index than what is in the database)

19:15:24 - 9
19:16:25 - 9
19:17:25 - 9
19:18:26 - 9

The other node reports a difference of 2 (2 less in the index than what is in the database)

18:28:16 - -2
18:29:17 - -2
18:30:17 - -2
18:31:18 - -2
18:32:18 - -2


What is the end user impact of a positive index count? What is an acceptable difference?

We see a similar occurrence for Zephyr indexes.

Right now these are the numbers:

{"dbCount":3324765,"indexCount":3324986,"nodeId":"1a-1","nodeIp":"1a-1"}
{"dbCount":3324765,"indexCount":3324759,"nodeId":"1c-1","nodeIp":"1c-1"}

These are generated by doing:

curl -s -u USERNAME 'localhost:8080/rest/zapi/latest/execution/indexCount'

The difference in the results of +221 and -6 respectively, what impact does this have to our end users? What behavior could they expect to see? 

[Matt Doar]

Well, one behaviour that the users might see is different results in Zephyr reports, depending on which Data Center node they are logged into. Are these just for Zephyr indexes or for Jira indexes as well?

I guess rebuilding the Zephyr index on one node offline would give you a reference number to compare other nodes to. Do the logs contain anything about sync failing between the Zephyr indexes?

I have seen the case where changing a remote link in an issue updates the database but not the regular Jira Lucene index. But I think that was fixed a while back.

But in general I've seen Jira Lucene indexes differ by a few hundred and not had user feedback about problems. Perhaps it just wasn't noticed!