Project

General

Profile

Actions

Feature #1131

closed

Add "load-warn-threshhold" for fastcgi settings

Added by mike503 about 17 years ago. Updated almost 8 years ago.

Status:
Invalid
Priority:
Normal
Category:
mod_fastcgi
Target version:
ASK QUESTIONS IN Forums:

Description

Allows for a message to be shown without the need to enable full fastcgi debugging

Suggested message: "Warning: load threshhold of $threshhold reached. Consider setting a higher value for the number of FastCGI engines." (and include additional data like binary, maybe vhost that triggered it, etc? - things to help debug/ensure the right one will be modified)

Right now the only way to measure fastcgi load is using fastcgi.debug = 1 and having the error log grow infinitely large and have to manually check it periodically for "load: XX" to be close to what i've set for the number of processes to spawn-fcgi.

FOR 1.5.x:

I'm looking at the docs now.

I'm thinking this setting should be related to the "proxy-core.max-pool-size" setting. I don't know how 1.5.x works, but it would be nice to be notified when a certain (configurable) threshhold is reached, so it is understood to change the configuration to start more engines.

Actions #1

Updated by mike503 about 17 years ago

Is this understandable? Does it sound like something that could be done?

Just wondering. It would be nice to have a fully automated true adaptive spawning engine (could one be packaged/integrated with lighty, somehow?) - especially now that even the existing on in 1.4.x will not be there in 1.5.x.

If I get a chance to try to hack the code and see where it prints out "load: XX" and see if I can add in a config setting, I'll be happy to upload a patch. I'm thinking this would be very simple for someone to implement. Remember, it's a way to get a notice in the error log without having to enable the full fastcgi.debug two-lines-per-fcgi-request noisy debugging logging.

Actions #2

Updated by mike503 about 17 years ago

okay, i finally installed 1.5.x to see if i could figure it out easily enough.

looks like there is a .debug=1 setting for each backend.

however, if it's not enabled, i can completely hammer the server (way more than the concurrent amount of requests defined in PHP_FCGI_CHILDREN and the .max-pool-size) and i do not get any warnings. in fact, just holding down F5 in IE it eventually causes the server to stop responding. with debugging on, it looks like it starts throwing requests into the mod_proxy_core_backlog and that might be when it stops responding.

i don't necessarily care at the moment to debug why it stops responding. i would still like a way to add some sort of proxy-core.warn-threshhold configuration that will report to the error log (with .debug NOT enabled) - or perhaps a trigger so it could be used to email "hey, you're starting to hit the threshhold, reconfigure your $proxy engines to spawn more"

Actions #3

Updated by gstrauss almost 8 years ago

  • Description updated (diff)
  • Assignee deleted (jan)
  • Priority changed from High to Normal
Actions #4

Updated by gstrauss almost 8 years ago

  • Status changed from New to Invalid

Right now the only way to measure fastcgi load is using fastcgi.debug = 1

That is not true. See mod_status https://redmine.lighttpd.net/projects/lighttpd/wiki/Docs_ModStatus

Actions

Also available in: Atom