Skip to content

Fix eos report log fields

Problem

The EOS report lines are lacking important fields, as explained in https://gitlab.cern.ch/cta/operations/-/issues/1612.

  1. Missing file size in the EOS report message &sec.app=tape_evict. This is important to identify 0-length files.
log=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx&path=/eos/ctacms/archive/archive/cms/ab600da08330e7f6dbe6a9696aaf5a8a4b9a5aa2da0ecb3c25fe58f55637f5a2/shuaiy/STARlight/officalMCRequest/condor_projects/log_CohJpsi/run_2232.err&ruid=2&rgid=2&td=daemon.2083468:812@eosctafst0124&host=&ts=1739145600&tns=456605456&evictcounter=&fileremoved=&error=no_tape_replica&sec.app=tape_evict

Alternatively, we may not need to print the file when there is no disk replica.

  1. Initial user can be missing. For example, the following prepare evict request was done by user ruid=109701&rgid=2766 but the log line ruid=0&rgid=0 mistakenly atributes it to root.
log=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx&path=/eos/ctacms/archive/cms/store/data/Run2024E/ParkingVBF7/RAW/v1/000/381/070/00000/2cf37ce0-471d-4934-9df0-4c872b438ce5.root&ruid=0&rgid=0&td=service@localhost&host=&ts=1739756899&tns=696613637&evictcounter=2&fileremoved=false&error=&sec.app=tape_evict
log=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx&path=/eos/ctacms/archive/cms/store/data/Run2024E/ParkingVBF7/RAW/v1/000/381/070/00000/2cf37ce0-471d-4934-9df0-4c872b438ce5.root&ruid=109701&rgid=2766&td=nobody@unknown&host=&ts=&tns=&event=evict&reqid=&reqcount=&reqlist=&evictcounter=&ondisk=true&ontape=true&firstprepare=&senttocta=&activity=&error=&sec.app=tape_wfe

We need to report the correct issuing user.

  1. Sometimes (as shown above) the ts=&tns=& timestamp is not reported.

  2. Some other scenarios:

  • eviction by user is reported with ruid=0 rgid=0 (wrong)
  • eviction by fstgcd is reported with ruid=2 and rgid=2 (why not..., but why?)
  • eviction by evict_dual.sh is reported with ruid=0 and rgid=2 (to avoid spreading bad practices further)

What is the expected correct behaviour?

The EOS report log lines described above should show the correct values, as explained.

Relevant logs and/or screenshots

Possible causes