Details
-
New Feature
-
Resolution: Fixed
-
Minor
-
None
-
None
-
9223372036854775807
Description
In DDN-3356, by Andreas Dilger:
We don't have a tool to do this today, but it would make sense to write a simple tool "lljobstat" to show the top jobs on a server in order to simplify debugging of high load problems, since this is a reasonably frequent request.
It should be included with the base Lustre RPMs, so it must not have any complex external dependencies that are not included in the base OS distro (el7, el8, sles15, ubuntu22).
It should read all of the local "..job_stats" files (by default, or --ost or --mdt, or a specific jobstats file if given as an argument) every 10s interval (configurable, either "-i N" or last argument) and prints the top e.g. 5 jobs (configurable "-c N"), one line per job similar to "iostat -x -k -z 10". It should show something useful when run with minimal arguments (eg. just the interval), so that users can use it to easily determine which jobs are driving the most load.
Since the job_stats has a large number of stats, it is not possible to fit all of them in a single 80-column line, so any operations that have samples = 0 should not be shown. Priority for display should be to show read, write (counts, if non-zero), read_bytes, write_bytes (in MiB/s units, if non-zero), then the top metadata ops by count. It probably makes sense to use abbreviations for the names, like llobdstat so that more can fit onto the line (cx: create, dx: destroy, st: statfs, pu: punch, etc). In the newer llstat and llobdstst it checks if the terminal width is over 80 and shows more fields, but this doesn't have to be in the first version.
To determine the "top" jobs, it probably makes sense to sum the operations for the same job name across all watched job_stats files, then sort by total count of operations (read+write, but not bytes) and include this as the second item shown ("ops: N") after the job name ("job: name", with escaping/quoting if needed). The timestamp should be shown for each interval.
Given that the input is YAML, the output could also be YAML, but only if it can be formatted nicely for human readability (one line per job, no excessive quoting). The main users of this will be people, since monitoring tools will likely read and process all of the job_stats output directly.
Attachments
Issue Links
Activity
Bjoern, is there a way to "try" loading the libyaml-dev CLoader, but fall back to the regular Loader if it is not installed?
See link for solution:
https://github.com/yaml/pyyaml/issues/108#issuecomment-370459912
Essentially libyaml-dev is missing on your system. It is required for the CLoader (which replaces the slow Python loader)
Greetings, Bjoern
bolausson, I pushed the "simple" version of your patch but it is reporting an error:
This is causing test failures:
lljobstat -n 1 -i 0 -c 1000
Traceback (most recent call last):
File "/usr/bin/lljobstat", line 15, in
from yaml import CLoader as Loader, CDumper as Dumper
ImportError: cannot import name 'CLoader'
"Oleg Drokin <green@whamcloud.com>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/48888/
Subject: LU-16228 utils: add lljobstat util
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: e2812e877314bc101efdc5a235c7fae8f7424f96
It looks like the newly-added sanity.sh test_205e needs to add a version check for interop testing:
trevis-82vm3: sh: lljobstat: command not found
There is a version check in test_205d already.
"Feng Lei <flei@whamcloud.com>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/48888
Subject: LU-16228 utils: add lljobstat util
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 777dd8757b0a121daf22f275f7eeb5a2b00ea62f
Feng Lei, this looks mostly good. I would say that the comment is large enough that it shouldn't be printed each time, maybe just document the abbreviations in the man page or if "-h" is used. I would suggest "ln" for link (to match the command name).
The "top_jobs:" should have an underscore so it is a single word, even though I know YAML does not require this, since it makes parsing easier with scripts (eg. "awk '/keyname:/ { print $2 }'".
adilger Is such an output OK?
# ./lljobstat # Abbr.: # cr: create, op: open, cl: close, mn: mknod, lk: link, # ul: unlink, mk: mkdir, rm: rmdir, mv: rename, ga: getattr, # sa: setattr, gx: getxattr, sx: setxattr, st: statfs, sy: sync, # rd: read, wr: write, pu: punch, mi: migrate, fa: fallocate, # dt: destroy, gi: get_info, si: set_info, qc: quotactl, pa: prealloc, timestamp: 1665557039 top jobs: - touch.500: {ops: 6, op: 1, cl: 1, mn: 1, ga: 1, sa: 2} - rm.0: {ops: 6, cl: 2, ul: 1, rm: 1, ga: 1, st: 1} - chown.0: {ops: 3, ga: 2, sa: 1} - bash.0: {ops: 2, ga: 2} - mkdir.0: {ops: 2, mk: 1, st: 1}
No, the time should be the current Unix timestamp in seconds:
# lctl get_param llite.*.stats llite.testfs-ffff89b1b9c27000.stats= snapshot_time 1665476432.161461498 secs.nsecs ioctl 502 samples [reqs] getattr 290 samples [usec] 56 1059 48623 11761597 getxattr 2 samples [usec] 975 30159 31134 910515906 inode_permission 298 samples [usec] 61 566 52783 11517621 opencount 295 samples [reqs] 1 1 295 295 # date +%s 1665476439
there is a bug on master that the timestamp is incorrectly printing the boot-relative time instead of the wallclock time. See LU-16231.
Yes this is possible with a try - except construct.
The CLoader worked perfeclty fine on default EXAScaler 5.2.7 install.
It was not neccessary to install any aditional packages except:
By the way, I added my enhneced version to the DDNeu GitHub repo:
https://github.com/DDNeu/global-lustre-jobstats
Cheers,
Bjoern