Unterminated string error

[shaftoe]

Hi folks, we've got Zenoss Core 3.2.0 and Postgres zenpack working OK, but we frequently get this kind of event alerts:

Device: db-master Component: Severity: Error Time: 2011/11/10 23:10:13.000 Message: postgres failure: Unterminated string starting at: line 1 column 8177 (char 8177)

Clear message comes always 5 minutes later:

Event: 'postgres failure: Unterminated string starting at: line 1 column 8177 (char 8177)' Cleared by: 'postgres connectivity restored'

After Zenpack install, we've started to get many of those messages into postgres log file: LOG: unexpected EOF on client connection

Googling around seems to be a client problem:

http://postgresql.1045698.n5.nabble.com/LOG-unexpected-EOF-on-client-connection-td3204658.html

One suggestion is to tweak tcp_keepalive settings for the client.

[cluther]

I updated this particular event to be a debug severity so hopefully it's less annoying. Let me know if you think there's a better way to handle it. The change is in the 1.0.2 version of the ZenPack that I just released.

[shaftoe]

Hi Luther, I'm sorry to bother you again with this issue, but we're still getting "Unterminated string starting at" alerts even with the latest version (1.0.2). Let me know if I can help to debug this. For the records, we've got Zenoss collector in US monitoring a Postgres server in EU.

[cluther]

@shaftoe, I thought I had responded to your comment, but looking back through issues I see that I only reopened the issue.

I think what happened is that I set the zEventSeverity on the postgresFailure mapping inside the /Status event class to error instead of original. So the lowered severity for this error that's coming in is being overridden by the mapping. You can verify this by going into that postgresFailure mapping in the /Status event class and changing zEventSeverity under its configuration properties to original instead of error.

Let me know if this resolves the problem. I'll update the ZenPack to set this by default.

[shaftoe]

@cluther , thanks as usual for your feedback. In the meanwhile we migrated our Zenoss Core in the same datacenter where the PostgreSQL lives, so hopefully we'll see a much lower rate of problems due to connectivity issues. I'll keep this issue updated though and follow your suggestion if in need.

[cluther]

@shaftoe, sounds good. I just packaged up version 1.0.3 with zEventSeverity on that mapping set to original by default to address for for anyone else who might run into the same problem in the future.

[shaftoe]

@cluther Unfortunately seems that the problem persists even with v1.0.3 (I'm referring to the "LOG: unexpected EOF on client connection" lines into postgresql log). Now we've got an average 2.7 milliseconds ping time between Zenoss and the DB box so I tend to exclude network connectivity problems. By the way, I've followed your suggestion and lowered the zEventSeverity event value to warning so we don't get email alerts anymore.