summaryrefslogtreecommitdiff
path: root/debian
diff options
context:
space:
mode:
authorGuillaume Friloux <guillaume.friloux@gmail.com>2014-02-10 16:01:40 +0100
committerGuillaume Friloux <guillaume.friloux@gmail.com>2014-02-10 16:01:40 +0100
commit42592e472a67e5ae5f399cb9ebc8abdf5e22a0cb (patch)
tree5470ae8375f679d8652fbc8158ea627ff1eb40ac /debian
parent190d34beb4495ace579333ab664329c210eebda3 (diff)
Fix an infinite loop if elasticsearch goes crazy.
We had a case here where elasticsearch stopped answering to HTTP queries SMMAN was sending. This led SMMAN to a state where he had 1015 sockets opened to elasticsearch, and constantly trying to open logfiles as they had new logs in it, but unable to open them because we reached the 1024 nofile limit. So this patch change behavior : - Add a 10s timeout to ecore_con_url objects - Dont instantly stat() files right after processing if an error happenned at last try. This bug happenned twice here with a server sending over 1k logs in the same second.
Diffstat (limited to '')
0 files changed, 0 insertions, 0 deletions