Multivac | Multivac
Check out the new release policy


It collects all troubles of CI.


  • Python 3

  • requests

How to use

Add a token on Personal access token GitHub page, give repo:public_repo access and copy the token to token.txt.

Collect logs:

$ ./multivac/ --branch master tarantool/tarantool
$ ./multivac/ --branch 2.8 tarantool/tarantool
$ ./multivac/ --branch 2.7 tarantool/tarantool
$ ./multivac/ --branch 1.10 tarantool/tarantool

If something went wrong during initial script run, you may re-run it with --nostop option: it disables stop heuristic. The heuristics is the following: stop on a two weeks old workflow run stored on a previous script call.

Generate report:

$ ./multivac/ --branch master --branch 2.8 --branch 2.7 --branch 1.10

Add --format html to get the ‘last seen’ report in the HTML format instead of CSV. Reports are stored in the output directory.

Caution: Don’t mix usual calls with --nologs calls (see below), otherwise some logs may be missed. The script is designed to either collect meta + logs or just meta. If the meta is up-to-date, there is no cheap way to ensure that all relevant jobs are collected with logs.

Time spent in jobs

Collect jobs metainformation:

$ ./multivac/ --nologs --nostop tarantool/tarantool

You may need to re-run it several times to collect enough information: GitHub ratelimits requests to 5000 per hour.

If hit by 403 error, look at the last X-RateLimit-Reset value in debug.log: it is time, when you may start the script again. Call date --date=@<..unix time..> '+%a %b %_d %H:%M:%S %Z %Y' to translate this value into a human readable format.

You may continue from a particular page using the --since N option (beware of holes, always leave some overlap).

Next, generate the report itself:

$ ./multivac/ [--short]

It prints minutes splitted in two ways:

  • per day / per week / per month

  • by ‘runs-on’ (‘ubuntu-20.04’ and so on)

Use --short to merge ‘ubuntu-18.04’ and ‘ubuntu-20.04’ into just ‘ubuntu’.

Found what you were looking for?