diff options
| author | Yevhenii Shapovalov <yshapova@redhat.com> | 2018-02-19 12:50:53 +0200 |
|---|---|---|
| committer | Yevhenii Shapovalov <yshapova@redhat.com> | 2018-02-19 12:52:38 +0200 |
| commit | 9bbb3de7d2c19922a19a61fcb4b2db68e8184fcc (patch) | |
| tree | d0fe1bb442e33b595c9b9b23f6dac07492f4143e /tests/multipathd_oom/PURPOSE | |
| parent | 5bea53fe7ed7dd7afa8e76f7f84f76fdb89d59d9 (diff) | |
| download | device-mapper-multipath-add_tests.tar.gz device-mapper-multipath-add_tests.tar.xz device-mapper-multipath-add_tests.zip | |
Diffstat (limited to 'tests/multipathd_oom/PURPOSE')
| -rw-r--r-- | tests/multipathd_oom/PURPOSE | 14 |
1 files changed, 14 insertions, 0 deletions
diff --git a/tests/multipathd_oom/PURPOSE b/tests/multipathd_oom/PURPOSE new file mode 100644 index 0000000..fbeab48 --- /dev/null +++ b/tests/multipathd_oom/PURPOSE @@ -0,0 +1,14 @@ +the oom_adj is -17 and oom_score_adj is -1000 after installing device-mapper-multipath-0.4.9-53.el6.x86_64 and starting multipathd. +This means multipathd will never be selected to be killed by oom according to the kernel doc. + + +Here are the steps to reproduce this issue. +1. create a multipathed disk /dev/mapper/mpathb via scsi_debug using all the free memory. this will consume most of the free mem. +2. io stress on the multipath disk, this step is option. +#dd if=/dev/zero of=/dev/mapper/mpathb & +3. consuming the free mem +#for i in `seq 10000`; do array[$i]=$(</var/log/messages); done & +4. trigger the oom_killer manually +#while true; do echo f > /proc/sysrq-trigger ; done +5. check the console or /var/log/message if multipathd is killed + |