Since 2013-12-10 we're seeing between 15-40 lines per day in the
sampled-1000 cache logs that have “nan” in the “Request service time”
column [1].
For the mobile-sampled-100 cache logs it is 10-30 such lines per day [2].
The zero logs do not show any affected lines up to now.
The affected lines come with “-/” for the “Request status / HTTP
status code” column, and “Reply size” is “-”.
[1] The first occurrence for the sampled-1000 stream can be obtained by
running
zgrep 49094481570 /a/squid/archive/sampled/sampled-1000.tsv.log-20131210.gz
on stat1002. That line is for cp1050.eqiad.wmnet on
2013-12-10T03:36:04.
[2] The first occurrence for the mobile-sampled-100 stream can be
obtained by running
zgrep 9650126710 /a/squid/archive/mobile/mobile-sampled-100.tsv.log-20131210.gz
on stat1002. That line is for cp1047.eqiad.wmnet on
2013-12-10T02:01:24.
Version: unspecified
Severity: normal