[LU-15826] jobstats output can produce invalid yaml ( part 2 )

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Affects Version/s: Lustre 2.12.6
• Severity: 3
• Assignee: WC Triage
• Reporter: James Beal

Description

The work in LU-13857 looks good, however we also see jobstat names which break the yaml for example

 

 - job_id: kworker/u515:5.
   snapshot_time: 1651751372

Which should be

 - job_id: "kworker/u515:5."
   snapshot_time: 1651751372

Issue Links

• is related to: LU-17358 jobstats output can produce invalid yaml ( part 3 )
• is related to: LU-13857 jobstats output can produce invalid yaml
• is related to: LU-11407 Improve stats data
• is related to: LU-16110 Make output of jobs_stats and rename_stats valid YAML

Comments

[Andreas Dilger]

The invalid characters in JobID issue was fixed with patch https://review.whamcloud.com/37764 "LU-11407 tgt: cleanup job_stats output printing".

The incorrect job_stas YAML formatting was fixed with patch https://review.whamcloud.com/48417 "LU-16110 lprocfs: make job_stats and rename_stats valid YAML".

[Alexander Boyko]

 adilger, It is not directly related to description, but LU-11407 lost indentation for yaml during refactoring.

- job_id:          fio.0
snapshot_time   : 42696.868451672 secs.nsecs
start_time      : 1384.907433838 secs.nsecs
elapsed_time    : 41311.961017834 secs.nsecs
  read_bytes:      { samples:    26054605, unit: bytes, min:     4096, max:  4194304, sum:   53228884680704, sumsq: 4159114415330820096 }
  write_bytes:     { samples:    12962712, unit: bytes, min:    65536, max:  4194304, sum:   34331481341952, sumsq: 2486223057455153152 } 
 

vs 2.12

- job_id:          fio.0
  snapshot_time:   1661494584
  read_bytes:      { samples:    30567346, unit: bytes, min:   65536, max: 4194304, sum:  48407357423616 }
  write_bytes:     { samples:    14357818, unit: bytes, min:   65536, max: 4194304, sum:  32058573324288 }

[James Beal]

Reviewed and commented.

 

We only have exascaler servers right now so I can't easily test, sorry.

[Andreas Dilger]

I thought this issue was fixed, but I see that patch https://review.whamcloud.com/37764 "LU-11407 tgt: cleanup job_stats output printing" has slipped through the cracks and did not land in 2.15.0.

If you are able to test and or review the patch it will speed up landing.