Re: [pinba] Persistence
20 views
Skip to first unread message
Antony Dovgal
Apr 22, 2013, 3:48:03 AM4/22/13
to pinba-...@googlegroups.com
On 2013-04-19 14:50, Iain Mckay wrote:
> Hi guys, is it normal behaviour for the collected stats not to persist across restarts of MySQL? If so, what do people here do about archiving them?
Sure, they're stored in the memory.
Archiving them? The main ideas is to have realtime data on the latest X seconds.
Why would anyone need to archive it?
It would also make reading/writing the data extremely slow, not to mention the giant storage required..
--
Wbr,
Antony Dovgal
> Hi guys, is it normal behaviour for the collected stats not to persist across restarts of MySQL? If so, what do people here do about archiving them?
Sure, they're stored in the memory.
Archiving them? The main ideas is to have realtime data on the latest X seconds.
Why would anyone need to archive it?
It would also make reading/writing the data extremely slow, not to mention the giant storage required..
--
Wbr,
Antony Dovgal
Iain Mckay
Apr 25, 2013, 8:33:16 AM4/25/13
to pinba-...@googlegroups.com
We want to archive the data for two reasons:
1. To pinpoint what pages misbehave and some useful information on where they are misbehaving (certain operations taking too long/too much memory).
2. To provide historical performance data.
The storage requirements are a non-issue for us we just wanted to know if there was a way to persist the data. Either by storing it in one massive table (with the performance issues you noted) or by archiving them to another longer term table where performance isn't an issue.
1. To pinpoint what pages misbehave and some useful information on where they are misbehaving (certain operations taking too long/too much memory).
2. To provide historical performance data.
The storage requirements are a non-issue for us we just wanted to know if there was a way to persist the data. Either by storing it in one massive table (with the performance issues you noted) or by archiving them to another longer term table where performance isn't an issue.
Antony Dovgal
Apr 25, 2013, 8:56:57 AM4/25/13
to pinba-...@googlegroups.com
On 2013-04-25 16:33, Iain Mckay wrote:
> We want to archive the data for two reasons:
>
> 1. To pinpoint what pages misbehave and some useful information on where they are misbehaving (certain operations taking too long/too much memory).
> 2. To provide historical performance data.
We do either of those using RRD - we have a lot of graphs here.
> We want to archive the data for two reasons:
>
> 1. To pinpoint what pages misbehave and some useful information on where they are misbehaving (certain operations taking too long/too much memory).
> 2. To provide historical performance data.
Of course, if there is some kind of data that isn't present on some graph, it's lost and cannot be restored.
> The storage requirements are a non-issue for us we just wanted to know if there was a way to persist the data. Either by storing it in one massive
>table (with the performance issues you noted) or by archiving them to another longer term table where performance isn't an issue.
--
Wbr,
Antony Dovgal
---
http://pinba.org - realtime profiling for PHP
Iain Mckay
Apr 25, 2013, 9:47:22 AM4/25/13
to pinba-...@googlegroups.com
We've done that in the past. Cron'd a script that copied live data in to an archive to keep read/writes nice and fast.
Thanks very much for your help.
Reply all
Reply to author
Forward
0 new messages