EndPointCorp / end-point-blog

End Point Dev blog
https://www.endpointdev.com/blog/
17 stars 65 forks source link

Comments for PostgreSQL log analysis / PGSI #488

Open phinjensen opened 6 years ago

phinjensen commented 6 years ago

Comments for https://www.endpointdev.com/blog/2011/08/postgresql-log-analysis-pgsi/ By Greg Sabino Mullane

To enter a comment:

  1. Log in to GitHub
  2. Leave a comment on this issue.
phinjensen commented 6 years ago
original author: Robert Treat
date: 2011-08-20T00:14:07-04:00

I like PGSI, and would like to replace a bunch of our pgfouine setups with it, but I think the variety of reports it produces are still below that of pgfouine. Yeah, I know, we could submit patches, but we've not hit enough of a tipping point to move to pgsi rather than tweak pgfouine as needed. That said, are there any plans for expanding the reports given?

phinjensen commented 6 years ago
original author: Greg Sabino Mullane
date: 2011-08-20T17:49:27-04:00

Good news, Rob, I just added PGSI to bugzilla, so if you are not up to patching, please at least add a feature request! :) http://bucardo.org/bugzilla/enter_bug.cgi?product=PGSI

phinjensen commented 6 years ago
original author: Michael Banck
date: 2011-08-23T08:33:10-04:00

One problem pgfouine has to my knowledge is parsing logs written in a non-english locale. Can PGSI deal with this?

phinjensen commented 6 years ago
original author: Greg Sabino Mullane
date: 2011-08-23T10:26:18-04:00

I don't see why PGSI would not be able to deal with other locales, but it would help to know what problems pgfouine runs into. Feel free to raise a bug if you find PGSI has any locale problems and we'll get it fixed. :)

phinjensen commented 6 years ago
original author: Michael Banck
date: 2011-08-23T11:01:21-04:00

AIUI, pgfouine parses the log file and stores the information. E.g. in a german installation, queries get logged like "LOG: Dauer: 3.080 ms Anweisung: SELECT [...]". I assume some mapping is required here in order to parse keywords correctly, and pgfouine was missing that last time I looked (a long time ago). Maybe PGSI is doing things differently so this is not a concern.

phinjensen commented 6 years ago
original author: Greg Sabino Mullane
date: 2011-08-23T11:55:44-04:00

Ah, okay. Yes, 'duration' is hard-coded in there for PGSI but I'll make having it depend on LANG/LC* in my todo list.

phinjensen commented 6 years ago
original author: alvherre
date: 2011-08-23T23:10:19-04:00

For the translated stuff, I wonder if it would work to extract the translations directly from the .po or .mo files. That way you don't have to hardcode them in PGSI's source.

phinjensen commented 6 years ago
original author: Greg Sabino Mullane
date: 2011-08-23T23:33:23-04:00

Well, we could grab the strings from there, but they would still need to be hard-coded as there is no guarantee that the files would be available on the box running pgsi, so we can't rely on grabbing them on the fly. A quick glance at the PGSI source shows that 'duration' may be the only thing that needs translating anyway (other than some of the less important syntax strings used to flatten similar queries)