Details
-
Bug
-
Resolution: Fixed
-
Major
-
5.5.0
-
5.5.0-2211
-
Untriaged
-
Windows 64-bit
-
No
Description
How it was seen
This was noticed while running eventing tests on windows server pool.
1) Run eventing jobs on a set of windows servers(Some of the scripts might pass/fail)
2) Each tests resets the cluster before starting.
3) Once the tests are run, the nodes are returned back to server pool and other scripts pickup these server.
4) When the new tests run they fail in installation as these processes still are running and are accessing the couchbase installed directories.
Findings on the server
Couchbase installations fail on the server and we noticed lot of processes with .rbf extension.
See
When we clicked on "Go to details" on these individual processes they pointed to eventing-consumer.exe as the owner.
See
At the end of rebalance tests there were 92 eventing-consumers running(172.23.211.46)
C:\Users\Administrator>tasklist | FIND "eventing-consumer.exe" /c
|
92
|
|
C:\Users\Administrator>
|
Impact
This causes all the subsequent tests to fail after eventing tests as the installation fails
Workaround
Currently we are killing eventing-consumer's manually or rebooting the servers manually before installation. I will add a fix in testrunner to kill eventing-consumers before installation.
I had a discussion with Abhishek Singh regarding. We have identified the scripts/jobs that caused these failures.
Following are the node/job combinations for the same( Jobs might not have any data in couple of days)
The hosts are available for live debugging. Please let us know once you are done with these hosts.
172.23.211.46
|
http://qa.sc.couchbase.com/job/test_suite_executor/55672/consoleText
|
172.23.108.66
|
http://qa.sc.couchbase.com/job/test_suite_executor/56140/consoleText
|
Attaching cbcollect logs for 172.23.211.46.