ucx icon indicating copy to clipboard operation
ucx copied to clipboard

[DNM] verify integration rule 2

Open changchengx opened this issue 4 years ago • 3 comments

reserve branch to be used in future

changchengx avatar Jun 29 '21 00:06 changchengx

Can one of the admins verify this patch?

swx-jenkins3 avatar Jun 29 '21 00:06 swx-jenkins3

@changchengx i believe this PR does not release some IB resources which causes other tests on this machine failed on ALLOC_UAR this test allocated too many UAR regions:

[root@swx-rdmz-ucx-new-02 s]# grep uverbs /proc/2143/maps|awk '{print $6}'|sort|uniq -c
    347 /dev/infiniband/uverbs0
    250 /dev/infiniband/uverbs2
[root@swx-rdmz-ucx-new-02 s]# ps -ef|grep 2143
swx-azu+  2143  1492  1 16:47 ?        00:03:12 /scrap/azure/agent-09/AZP_WORKSPACE/2/s/build-test/test/gtest/gtest --gtest_filter=* -p 10 -i 2.0
[root@swx-rdmz-ucx-new-02 s]# cd /scrap/azure/agent-09/AZP_WORKSPACE/2/s
[root@swx-rdmz-ucx-new-02 s]# git branch
* (detached from pull/7018/merge)

Pls do not run any more tests on this PR before resolving the issue

[root@swx-rdmz-ucx-new-02 ~]# for p in `pgrep gtest`; do echo -n "$p "; cat /proc/$p/maps|grep uverbs|wc -l; done
2143 597
5013 59
6914 0
9533 58
15895 62
18098 76
18304 48
19336 68
24726 225
25873 58
32620 80

yosefe avatar Oct 12 '21 19:10 yosefe

@changchengx i believe this PR does not release some IB resources which causes other tests on this machine failed on ALLOC_UAR this test allocated too many UAR regions:

I've fixed the UAR resource release bug. I won't trigger CI for this PR if it's not quite necessary.

changchengx avatar Oct 13 '21 12:10 changchengx