[Linaro-validation] Old cache clean up on lava server
andy.doan at linaro.org
Thu Jun 7 20:43:31 UTC 2012
On 06/07/2012 02:43 AM, Spring Zhang wrote:
> Hi, all,
> There's still a WI in squid proxy blueprint. That is to clean up the old
> cache. I jumped into linaro cache directory on v.l.o, and found it cost
> about 100GB.
> control:/linaro/images/cache$ du * -sh
> 328M ci.linaro.org <http://ci.linaro.org>
> 36M people.linaro.org <http://people.linaro.org>
> 3.4G releases-linaro-org
> 10G releases.linaro.org <http://releases.linaro.org>
> 21G snapshots-linaro-org
> 63G snapshots.linaro.org <http://snapshots.linaro.org>
> control:/linaro/images/cache$ ll
> total 32
> drwxr-xr-x 8 root root 4096 2012-05-11 23:26 ./
> drwxr-xr-x 7 root root 4096 2012-01-25 01:16 ../
> drwxr-xr-x 3 root root 4096 2012-03-30 12:07 ci.linaro.org/
> drwxr-xr-x 3 root root 4096 2012-04-03 05:34 people.linaro.org/
> drwxr-xr-x 4 root root 4096 2012-05-24 16:27 releases-linaro-org/
> drwxr-xr-x 10 root root 4096 2012-05-09 14:31 releases.linaro.org/
> drwxr-xr-x 3 root root 4096 2012-05-11 23:26 snapshots-linaro-org/
> drwxr-xr-x 6 root root 4096 2012-05-02 18:55 snapshots.linaro.org/
> For some health check jobs, they need some cache. I think it's
> releases-linaro-org/ and snapshots-linaro-org/.
> So can we clean up others?
Clean up the others for now, but I actually just noticed a problem in
our current code. The only thing that we should now be managing cache
for are the root.tgz and boot.tgz files. However, we aren't caching
pre-built images properly. In master.py we are still calling
download_with_cache for pre-built images rather than downloading via the
I have a small local branch with a fix for this. I forgot about it, so
I'm going to test on our staging instance to make sure it hasn't
suffered from bit rot. Once that's fixed, I think we can deploy the fix,
and remove all non boot/root tarballs from the cache.
Does this make sense?
More information about the linaro-validation