schedpolicy from ubuntu_stress_smoke_test failed on D-AWS t2.small

[Po-Hsu Lin]

Reproduce rate: 3 out of 3 attempts
Instance: t2.small

This sub test is only failing on t2.small with D-aws (5.0.0-1022.25)

 schedpolicy STARTING
 schedpolicy RETURNED 2
 schedpolicy FAILED
 stress-ng: debug: [12598] 1 processor online, 1 processor configured
 stress-ng: info: [12598] dispatching hogs: 4 schedpolicy
 stress-ng: debug: [12598] cache allocate: default cache size: 30720K
 stress-ng: debug: [12598] starting stressors
 stress-ng: debug: [12598] 4 stressors spawned
 stress-ng: debug: [12602] stress-ng-schedpolicy: started [12602] (instance 3)
 stress-ng: debug: [12601] stress-ng-schedpolicy: started [12601] (instance 2)
 stress-ng: debug: [12600] stress-ng-schedpolicy: started [12600] (instance 1)
 stress-ng: debug: [12599] stress-ng-schedpolicy: started [12599] (instance 0)
 stress-ng: fail: [12600] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 stress-ng: fail: [12599] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 stress-ng: fail: [12600] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 stress-ng: fail: [12599] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 stress-ng: fail: [12600] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 stress-ng: fail: [12599] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 stress-ng: fail: [12600] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 stress-ng: fail: [12599] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 stress-ng: fail: [12600] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 info: 5 failures reached, aborting stress process
 stress-ng: debug: [12600] stress-ng-schedpolicy: exited [12600] (instance 1)
 stress-ng: fail: [12599] stress-ng-schedpolicy: sched_setscheduler failed: errno=16 (Device or resource busy) for scheduler policy deadline
 info: 5 failures reached, aborting stress process
 stress-ng: debug: [12599] stress-ng-schedpolicy: exited [12599] (instance 0)
 stress-ng: error: [12598] process 12599 (stress-ng-schedpolicy) terminated with an error, exit status=1 (stress-ng core failure)
 stress-ng: debug: [12598] process [12599] terminated
 stress-ng: error: [12598] process 12600 (stress-ng-schedpolicy) terminated with an error, exit status=1 (stress-ng core failure)
 stress-ng: debug: [12598] process [12600] terminated
 stress-ng: debug: [12602] stress-ng-schedpolicy: exited [12602] (instance 3)
 stress-ng: debug: [12601] stress-ng-schedpolicy: exited [12601] (instance 2)
 stress-ng: debug: [12598] process [12601] terminated
 stress-ng: debug: [12598] process [12602] terminated
 stress-ng: info: [12598] unsuccessful run completed in 5.01s

[Po-Hsu Lin (cypressyew)]

This issue can be found on instance a1.medium as well.

[Po-Hsu Lin (cypressyew)]

This is affecting GCP 5.0 Bionic (5.0.0-1026.27~18.04.1-gcp) on every instances.

[Po-Hsu Lin (cypressyew)]

Affecting GCP 5.3 Bionic (5.3.0-1009.10~18.04.1) as well.

[Po-Hsu Lin (cypressyew)]

Affecting Azure Disco 5.0 (5.0.0-1027.29-azure), but just instance Standard_B1ms.

[Colin Ian King (colin-king)]

EBUSY is a legitimate error to ignore for deadline scheduling on these kernels. Fix committed:

https://kernel.ubuntu.com/git/cking/stress-ng.git/commit/?id=3b56c2f3568b5b3ab4310962f6240cc7773b7d41

[Colin Ian King (colin-king)]

@Sam, when convenient, do you mind re-testing these. Thanks for reporting this.

[Po-Hsu Lin (cypressyew)]

Rested and passed on all the instances / kernels mentioned above.
Thank you!