|Subject:||Re: cache purge|
|Date:||Nov 16, 2009 11:33:17 pm|
In case it matters, here is the error I received in the error log:
2009/11/17 15:03:52 [crit] 1553#0: *1 cache file "/data/nginx_cache/etwebservice/a/2b/b3a2527b6f3a38d63663ee436e7d82ba" is too small, client: 188.8.131.52, server: localhost, request: "HEAD /purge/RestTakeoutServer/dish/B10I24R57547/list?commentCount=5 HTTP/1.1", host: "184.108.40.206"
One of my developers did the patch manually and she might have introduced some bugs during the process.
Would it be easier to make it a standalone module rather than a patch for users?
I'm waiting for your new patch to the latest nginx and I'll test it out and report back.
I'm investigating cache purge in nginx and currently we have some success with this patch: http://labs.frickle.com/nginx_cache_purge/. The reason I said "some" is the patch works normally but blows out (the message was like "the file to purge is too small...")when the nginx process has been restarted and the cache to purge was created by the previous nginx process.
It's a shame that you didn't report it, because I was unaware of this until now.
Anyway, I'm unable to reproduce this bug (purging files from cache after restarting nginx works as expected, at least with nginx-0.8.8, so maybe you introduced some bugs while syncing it on your own?)
Also, the message you are referring to ("cache file \"%s\" is too small") is produced by nginx when it's trying to open cached file (which is also done before purging it), so it looks like something is wrong... It would be great if you could produce debug log and send it either on- or off-list.
The patch is also a little out of of sync with the latest nginx.
Actually, I was going to sync it with the latest nginx and push the release yesterday, together with ngx_supervisord-1.1, but something came up and I didn't do it...
But I'm gonna do it in next few days (maybe even today), stay tuned! ;)
Best regards, Piotr Sikora < piot...@frickle.com >