[OSM-dev] dev server unresponsive
osm-list at deelkar.net
Wed Apr 25 17:01:31 BST 2007
-----BEGIN PGP SIGNED MESSAGE-----
Nick Hill schrieb:
> Hello Dirk
> The cronjob was updatedb which indexed the entire T at H tileset. I have
> pruned T at H from the updatedb process.
> There were more than 150 processes competing, most of which were apache
> processes. The apache settings were insane. I have corrected these.
> Don't know how settings were changed to 150 clients!
> Tile writing through T at H is saturating the T at H drive. This is impacting
> read performance for T at H. I guess tiles are being generated appx 30x
> faster than they are being read, and faster than the hard drive can
> cope. The T at H scheduler needs to pull back.
I added some rudimentary exponential backoff for the tile upload in case
it fails somewhere. I remember planning to add this way back when the
tiles were still stored in the db, but somehow never got around to
actually do it.
This needs tweaking, and probably upload.pl should remember failure
counts from previous runs so the backoff is actually exponential over a
chain of failures.
The problem today was, that probably the uploads went fine, but since
the hdd was overloaded it timed out, and the upload clients only saw a
failure, only to upload the same tileset over and over, adding to the
Well we shall see how this works under normal conditions with the backoff.
Dirk-Lüder "Deelkar" Kreie
Bremen - 53.0952°N 8.8652°E
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
-----END PGP SIGNATURE-----
More information about the dev