search

fgci-org / fgci-ansible

:microscope: Collection of the Finnish Grid and Cloud Infrastructure Ansible playbooks
MIT License
54 stars 18 forks source link

squid configuration re-assessment #143

Closed A1ve5 closed 8 years ago

A1ve5 commented 8 years ago

squid is mostly running on default configuration values. Cache size seems to be small. The fitness of its configurations should be re-assessed and if necessary fixed.

martbhell commented 8 years ago

For example the disk cache is not configured - it only stores the cache in memory.

https://github.com/CSC-IT-Center-for-Science/ansible-role-squid/blob/master/templates/squid.conf.j2#L72

A1ve5 commented 8 years ago

cache_dir gets added but still something isn't optimal with these confs

martbhell commented 8 years ago

You're right.

martbhell commented 8 years ago

a)

Maybe having two cache_dirs is bad (for a small cluster) - as the requests are load-balanced between the caches (based on load by default). I took the smaller/default cache away and on the second reinstall of a compute node the cache was used quite a lot. Maybe we can remove the 100MB 1 cache? https://github.com/CSC-IT-Center-for-Science/ansible-role-squid/blob/master/templates/squid.conf.j2#L87

b) the extra cache is 100000 (100GB) - but / is only 50GB on all the clusters.

cache_dir ufs /squid/cache 100000 16 256

martbhell commented 8 years ago

Putting this in review. Let's review in a few weeks (when this change hopefully has propagated to the clusters).

martbhell commented 8 years ago

Adding to requirements - closing. Let's open a new issue if we want to improve more.