Fail2ban-0.9.1 segafults under high I/O and cpu load. The error message is "fail2ban-server[4061]: segfault at 40 ip 00007fa5c776b290 sp 00007fa5af7fb6c8 error 4 in libsqlite3.so.0.8.6[7fa5c774d000+bb000]" This has been fixed upstream https://github.com/fail2ban/fail2ban/issues/844 with a fix at https://github.com/fail2ban/fail2ban/pull/839/files
I also got the problem, 10 of my 21 servers have lost their fail2ban process in a period of a month. Since, I added a script into each server's cron to check process (and respawn if needed) every hours until fixed...
I'm pretty sure upstream previously said 0.9.1 was deemed stable. Going back to 0.8.14 would mean a significant loss of new features and general improvements, but it's the exact reason I left 0.8.* in the tree to begin with.
Yes, it's probably "stable" on a test server that does nothing special :-/ I noticed that it mostly crashed during server backup time, the backup is a client side rsync launched by backuppc so under heavy I/O and CPU load...
0.9.2 is in the tree. Could you test that version?
(In reply to Jeroen Roovers from comment #4) > 0.9.2 is in the tree. Could you test that version? Yes, I've started to update all my servers today. I'll give you the result before the end of May...
I had fail2ban segfaults on multiple servers several times a day. Yesterday I updated them to 0.9.2 and this morning I had one segfault left. Then I noted that on that particular server I had forgot the update. So I'd say this is very strong evidence that 0.9.2 fixes the issue :-) Haven't seen a segfault on any of the machines where 0.9.2 was running. If others can confirm that this fixes the issue I'd suggest we try to stabilize 0.9.2 asap.
I have also updated to 0.9.2 and have not get any segfaults anymore. For me it looks also, that this version solves the issue.