integration/aof test unit fails on fedora 24 docker image

[melissop]

Hi,

I try to run the test suite on a fedora 24 docker image and integration/aof suite consistently fails. To be more specific all "assert_equal 1 [is_alive $srv]" statements in tests/integration/aof.tcl fail.
I tried to debug start_server_aof function but my tcl knowledge is limited.

I experimented with ubuntu and centos 7 images also without problems.
Tests conducted over commit 6712bce

               The End

Execution time of different units:
2 seconds - integration/aof

!!! WARNING The following tests failed:

*** [err]: Unfinished MULTI: Server should start if load-truncated is yes in tests/integration/aof.tcl
Expected '0' to be equal to '1'
*** [err]: Short read: Server should start if load-truncated is yes in tests/integration/aof.tcl
Expected '0' to be equal to '1'
*** [err]: Short read + command: Server should start in tests/integration/aof.tcl
Expected '0' to be equal to '1'
*** [err]: Fixed AOF: Server should have been started in tests/integration/aof.tcl
Expected '0' to be equal to '1'
*** [err]: AOF+SPOP: Server should have been started in tests/integration/aof.tcl
Expected '0' to be equal to '1'
*** [err]: AOF+SPOP: Server should have been started in tests/integration/aof.tcl
Expected '0' to be equal to '1'
*** [err]: AOF+EXPIRE: Server should have been started in tests/integration/aof.tcl
Expected '0' to be equal to '1'
Cleanup: may take some time... OK

[stevelipinski]

Having same problem on RH8 docker image when building 6.0.1. Did you get anywhere with this?
Best I can tell its related to the container environment (happens on one K8s build cluster, but not another). Maybe timing or security? Or the backing-volume? Just tossing out ideas.

[stevelipinski]

Discovered that my issue was there was no 'ps' command in my docker image - the is_alive proc uses ps -p <pid>.
My solution was that I needed to yum install procps beforehand.
In case that helps anyone...