[LU-12223] interop: conf-sanity test 28a fails with 'check lustre-OST0000.ost.writethrough_cache_enable failed!'

• Status: Open
• Project: Lustre
• Affects Version/s: Lustre 2.12.1, Lustre 2.14.0, Lustre 2.12.4, Lustre 2.12.6
• Type: Bug
• Priority: Minor
• Reporter: James Nunez (Inactive)
• Assignee: WC Triage
• Resolution: Unresolved
• Labels: interop

Environment

2.12.x servers and 2.10.7 and 2.11.0 clients for x > 0.100 with ldiskfs

• Severity: 3

Description

conf-sanity test_28a started failing on 2019-04-21 for 2.12.x servers and 2.10.7 and 2.11.0 clients.

Looking at a recent failures between 2.12.1 RC1 servers and 2.10.7 clients, https://testing.whamcloud.com/test_sets/6d0d78a8-6622-11e9-aeec-52540065bddc , we try to get/set the param obdfilter.lustre-OST0000.writethrough_cache_enable, but fail. From the client test_log

CMD: trevis-35vm11 /usr/sbin/lctl get_param -n obdfilter.lustre-OST0000.writethrough_cache_enable
trevis-35vm11: error: get_param: param_path 'obdfilter/lustre-OST0000/writethrough_cache_enable': No such file or directory
CMD: trevis-35vm11 /usr/sbin/lctl get_param -n obdfilter.lustre-OST0000.writethrough_cache_enable
trevis-35vm11: error: get_param: param_path 'obdfilter/lustre-OST0000/writethrough_cache_enable': No such file or directory
Setting lustre-OST0000.ost.writethrough_cache_enable from  to 1
CMD: trevis-35vm12 /usr/sbin/lctl conf_param lustre-OST0000.ost.writethrough_cache_enable='1'
CMD: trevis-35vm11 /usr/sbin/lctl get_param -n obdfilter.lustre-OST0000.writethrough_cache_enable
trevis-35vm11: error: get_param: param_path 'obdfilter/lustre-OST0000/writethrough_cache_enable': No such file or directory 
…
CMD: trevis-35vm11 /usr/sbin/lctl get_param -n obdfilter.lustre-OST0000.writethrough_cache_enable
trevis-35vm11: error: get_param: param_path 'obdfilter/lustre-OST0000/writethrough_cache_enable': No such file or directory
Update not seen after 90s: wanted '1' got ''
 conf-sanity test_28a: @@@@@@ FAIL: check lustre-OST0000.ost.writethrough_cache_enable failed!

For 2.12.0.94 servers, we didn’t see this issue; https://testing.whamcloud.com/test_sets/dc4a37b2-61de-11e9-92fe-52540065bddc
The first time we saw this failure was for 2.12.0.100 servers; https://testing.whamcloud.com/test_sets/9065a4c0-64d2-11e9-aeec-52540065bddc

Issue Links

• is related to: LU-11110 Remove obdfilter.*.* symlinks step by...

Comments

[James A Simmons]

Is this ZFS ? For ZFS writethrough_cache_enable doesn't exist.

[James Nunez (Inactive)]

Good question. This is for ldiskfs testing.

[Minh Diep]

test_28a() { # LU-4221
[[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
skip "Need OST version at least 2.5.52"
[ "$(facet_fstype ost1)" = "zfs" ] &&
skip "LU-4221: no such proc params for ZFS OSTs"

I see we skip on zfs

[Andreas Dilger]

The problem here is because conf-sanity test_28a on b2_10 is using parameter "obdfilter.*.writethrough_cache_enable" but this moved to "osd-ldiskfs.*.writethrough_cache_enable" and removed from conf-sanity in commit v2_11_54_0-77-g115bba9ffbe6. This is another case where testing with the old conf-sanity.sh from b2_10 should skip this subtest.