lp:~ballot/ubuntu-repository-cache/cache_swap

Branch merges

Propose for merging

No branches dependent on this one.

Merged into lp:ubuntu-repository-cache/trunk at revision 269

Haw Loeung: Approve (+1) on 2020-04-14

Laurent Sesquès: Approve on 2020-04-14

Canonical IS Reviewers: Pending requested 2020-04-14

Diff: 14 lines (+4/-0)

1 file modified

templates/squid-deb-proxy/squid-deb-proxy.conf (+4/-0)

api

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

268. By Benjamin Allot on 2020-04-14

Reduce the threshold of cache_swap_low and cache_swap_high

Try to reclaim space in the cache more aggrssively, as this is the
main suspected culprit in having an overall cache size getting
higher than the cache_dir size parameter.

For context, see http://squid-web-proxy-cache.1019090.n4.nabble.com/squid-3-5-27-does-not-respect-cache-dir-size-but-uses-100-of-partition-and-fails-td4685845.html#a4685883

267. By Benjamin Allot on 2020-04-14

Delete the old metadata items befoire syncing the new ones.

The rationale behind this change is to avoid filling the disks on the non-leader
peers.
The cache size is quite aggrssive (and rightly so) but we keep the current metadata
and the previous one.
On the leader node, the metadata are cleansed after the sync is done.
However, on the others nodes, the cache size might be a bit different, resulting
in reaching 100% space used and preventing the sync to be effective, breaking squid.
Since the leader is the source of truth, and we keep the previous good metadata repository,
we can safely remove before proceeding with the copy.

266. By Haw Loeung on 2020-02-04

[trivial] Fixed output on retry attempts

265. By Haw Loeung on 2020-02-04

Make metadata sync more resilient and retry

For Ubuntu Archive, it's fine because the sync host usually resolves
to more than one host/IP and therefore it goes through each but for
something like Ubuntu Ports, it's only a single host/IP. Instead,
let's retry up to 3 times before moving onto the next host.

Reviewed-on: https://code.launchpad.net/~hloeung/ubuntu-repository-cache/retry-for-each-host/+merge/378544
Reviewed-by: Paul Collins <email address hidden>

264. By Haw Loeung on 2019-11-25

Fix maximum_object_size_in_memory

Reviewed-on: https://code.launchpad.net/~hloeung/ubuntu-repository-cache/fix-maximum_object_size_in_memory/+merge/375940
Reviewed-by: Thomas Cuthbert <email address hidden>

263. By Haw Loeung on 2019-11-20

Ensure MPM tuning settings are applied, also graceful apache on change

Reviewed-on: https://code.launchpad.net/~hloeung/ubuntu-repository-cache/fix-apache2-mpm/+merge/375742
Reviewed-by: Thomas Cuthbert <email address hidden>

262. By Haw Loeung on 2019-09-09

Don't use hookenv.config() for apache root

ubuntu_repository_cache_sync() runs as the 'www-sync' user but permissions of charm/.juju-persistent-config as used by hookenv.config() are restricted to root 0600. Instead, we work out apache root from config file.

Reviewed-on: https://code.launchpad.net/~hloeung/ubuntu-repository-cache/no-hookenv-config-for-apache-root/+merge/372458
Reviewed-by: Joel Sing <email address hidden>

261. By Haw Loeung on 2019-09-09

Ship out config making it easier to trigger manual syncs - LP: #1834545

Reviewed-on: https://code.launchpad.net/~hloeung/ubuntu-repository-cache/cron-config/+merge/372457
Reviewed-by: Joel Sing <email address hidden>

260. By Laurent Sesquès on 2019-08-19

Update the check_http nrpe check to prepend and append slashes to base-path for the --url option (which nagios expects)

Reviewed-on: https://code.launchpad.net/~sajoupa/ubuntu-repository-cache/fix-check-http/+merge/371459
Reviewed-by: Jay Kuri <email address hidden>

259. By Haw Loeung on 2019-07-18

Allow overriding storage space used by squid cache

Reviewed-on: https://code.launchpad.net/~hloeung/ubuntu-repository-cache/allow-override-squid-cache-storage/+merge/369566
Reviewed-by: Tom Haddon <email address hidden>