[LU-16330] recovery-small test_152: QoS allocation slower than RR, killable semaphore doesn't work

• Status: Open
• Project: Lustre
• Type: Bug
• Priority: Minor
• Reporter: Maloo
• Assignee: WC Triage
• Resolution: Unresolved
• Severity: 3

Description

This issue was created by maloo for Dongyang Li <[e-mail]>

This issue relates to the following test suite run: https://testing.whamcloud.com/test_sets/6423df40-8be8-4ceb-9efc-e35a41855158

test_152 failed with the following error:

QoS allocation slower than RR, killable semaphore doesn't work

== recovery-small test 152: QoS object allocation could be awakened in case of OST failover ========================================================== 08:52:19 (1668761539)
CMD: trevis-60vm4 uname -r
striped dir -i0 -c1 -H crush2 /mnt/lustre/d152.recovery-small
striped dir -i0 -c1 -H fnv_1a_64 /mnt/lustre/d152.recovery-small/rr
striped dir -i0 -c1 -H crush2 /mnt/lustre/d152.recovery-small/qos
CMD: trevis-60vm4 /usr/sbin/lctl set_param fail_loc=0x80000173 fail_val=20
fail_loc=0x80000173
fail_val=20
CMD: trevis-60vm4 /usr/sbin/lctl get_param -n lov.*0000*.qos_threshold_rr
CMD: trevis-60vm4 /usr/sbin/lctl set_param lov.*.qos_threshold_rr=0
lov.lustre-MDT0000-mdtlov.qos_threshold_rr=0
lov.lustre-MDT0002-mdtlov.qos_threshold_rr=0
QoS allocation took 21 seconds
 recovery-small test_152: @@@@@@ FAIL: QoS allocation slower than RR, killable semaphore doesn't work 
  Trace dump:
  = /usr/lib64/lustre/tests/test-framework.sh:6532:error()
  = /usr/lib64/lustre/tests/recovery-small.sh:3403:test_152()
  = /usr/lib64/lustre/tests/test-framework.sh:6868:run_one()
  = /usr/lib64/lustre/tests/test-framework.sh:6918:run_one_logged()
  = /usr/lib64/lustre/tests/test-framework.sh:6755:run_test()
  = /usr/lib64/lustre/tests/recovery-small.sh:3405:main()
Dumping lctl log to /autotest/autotest-2/2022-11-18/lustre-reviews_review-dne-zfs-part-5_90735_16_7d42ba00-0350-47d9-88c8-13488e0d2b82//recovery-small.test_152.*.1668761581.log
CMD: trevis-60vm1.trevis.whamcloud.com,trevis-60vm2,trevis-60vm3,trevis-60vm4,trevis-60vm5 /usr/sbin/lctl dk > /autotest/autotest-2/2022-11-18/lustre-reviews_review-dne-zfs-part-5_90735_16_7d42ba00-0350-47d9-88c8-13488e0d2b82//recovery-small.test_152.debug_log.\$(hostname -s).1668761581.log;
		dmesg > /autotest/autotest-2/2022-11-18/lustre-reviews_review-dne-zfs-part-5_90735_16_7d42ba00-0350-47d9-88c8-13488e0d2b82//recovery-small.test_152.dmesg.\$(hostname -s).1668761581.log
CMD: trevis-60vm4 /usr/sbin/lctl set_param lov.*.qos_threshold_rr=17%
lov.lustre-MDT0000-mdtlov.qos_threshold_rr=17%
lov.lustre-MDT0002-mdtlov.qos_threshold_rr=17%

checked log, no seq rollover happening and seq is not used up for sync creation. Could not reproduce with the same env on local box. Slow disk?

VVVVVVV DO NOT REMOVE LINES BELOW, Added by Maloo for auto-association VVVVVVV
recovery-small test_152 - QoS allocation slower than RR, killable semaphore doesn't work