I've been working on a PGsql helper class to speed up the hotspots in my new pgsql PostgreSQL driver; so far it works rather well, and the performance of the new driver has leaped from 10% of the libpq module, to about 45% of the libpq module.
I won't be including the module until I have it at about 100% (or greater) of the libpq speed though.
Stephen R. van den Berg wrote:
I've been working on a PGsql helper class to speed up the hotspots in my new pgsql PostgreSQL driver; so far it works rather well, and the performance of the new driver has leaped from 10% of the libpq module, to about 45% of the libpq module.
Well, I'm pleased to announce that the current version is 62% *faster* than the old libpq-Postgres module in the worst case, and even faster on large blob queries.
Stephen R. van den Berg wrote:
Stephen R. van den Berg wrote:
I've been working on a PGsql helper class to speed up the hotspots in my new pgsql PostgreSQL driver; so far it works rather well, and the performance of the new driver has leaped from 10% of the libpq module, to about 45% of the libpq module.
Well, I'm pleased to announce that the current version is 62% *faster* than the old libpq-Postgres module in the worst case, and even faster on large blob queries.
And I (sadly) had to disable the module for now, because it still seems to be causing problems on Windows. Anyone capable of fetching me the error log?
It's double-sad, because the libpq version of the postgres driver doesn't seem to work on Windows either.
Press red blob of choise on http://pike.ida.liu.se/development/pikefarm/7.7.xml, then press "makelog.html", which will lead you to something like:
http://pike.ida.liu.se/generated/pikefarm/7.7/3169_448/makelog.html
Stephen R. van den Berg wrote:
And I (sadly) had to disable the module for now, because it still seems to be causing problems on Windows. Anyone capable of fetching me the error log?
Sorry for the noise. Figured out I could retrieve the error logs through the farm interface.
The script is not exporting again so Stephen never saw my answer. Can I have the script so I can set it up on the Pike computer instead?
Peter Bortas @ Pike developers forum wrote:
The script is not exporting again so Stephen never saw my answer. Can I have the script so I can set it up on the Pike computer instead?
Any other answers I missed? How about the poll/file_peek issue?
I can maintain it if it's running there, so it's a bad idea. And others depend om me running it, so I rather get notifications about when it's not working.
16663302 never exported it seems, but it looks like this:
Press red blob of choise on http://pike.ida.liu.se/development/pikefarm/7.7.xml, then press "makelog.html", which will lead you to something like:
http://pike.ida.liu.se/generated/pikefarm/7.7/3169_448/makelog.html (16663302) /Peter Bortas/---
It seems like it's the mailing server that was having problems due to high load/misconfiguration. I hope I fixed the problem now, but I'm not 100% sure. Hopefully the old emails will be dropping in later on.
I don't have admin-privs to that system, so no I can't.
I'm not so interested in keeping another system running either. I apparently have enough problems with this one.
Hedda (Lite ???) @ Pike (-) developers forum wrote:
It seems like it's the mailing server that was having problems due to high load/misconfiguration. I hope I fixed the problem now, but I'm not 100% sure. Hopefully the old emails will be dropping in later on.
I just received quite a large number of backlog mails (around 12, starting last Tuesday).
On Sat, Jul 26, 2008 at 04:50:03PM +0000, Hedda (Lite ���) @ Pike (-) developers forum wrote:
It seems like it's the mailing server that was having problems due to high load/misconfiguration. I hope I fixed the problem now, but I'm not 100% sure. Hopefully the old emails will be dropping in later on.
well, the mailserver having problems would be a good reason to run the script on the pike server.
i'd like to see the script though to have some sample code for interacting with lyskom. what language is it written in anyways?
greetings, martin.
Seems like the problem was that Lysator's email system was shut down for a while (16660679) and had to catch up. Not much to do about that, except running the email-list somewhere else.
pike-devel@lists.lysator.liu.se