length_in_sec value is too large

[ccabrera]

I have 2 different and completely unrelated clients, both with Vicidial version 2.14-617a BUILD: 170529-2320. In the past days, they have reported me that their statistics are off by millions of hours.

After analyzing their data, I've found that in both the vicidial_log and vicidial_closer_log table, the length_in_sec field is really, really off the expected value. However, both end_epoch and start_epoch are fine.

Sample values look like these (start, end, length):

start_epoch end_epoch length_in_sec
1497290813 1497291837 1497290813
1497291837 1497292019 1497290947
1497366374 1497366766 1497365113
1497371131 1497371606 1497369864
1497374500 1497374715 1497372909
1497375008 1497375373 1497373538
1497390024 1497390814 1497174868
1497392069 1497392532 1497371164
1497456619 1497456968 1497454937

From what I can notice, the length_in_sec is getting a unix timestamp instead of the (end - start) difference. This only happens in SOME entries. Most of them are fine, but with only 1 wrong value, any other correct value becomes useless.

Any ideas?

[mflorell]

This bug was fixed over the weekend, so you'll need to upgrade to the latest svn/trunk to get the fix.

Also, there is a new flag added to the AST_sleanup_agent_log.pl script to fix these log records(--check-call-lengths), so after upgrading, you'll want to run this on your servers:

/usr/share/astguiclient/AST_cleanup_agent_log.pl --check-call-lengths --last-24hours
/usr/share/astguiclient/AST_cleanup_agent_log.pl --check-call-lengths --last-30days