docker-cleanup.sh is missing from ansible, causing disk space to be exhausted
The docker-cleanup.sh script present on old non-ansible-generated build slaves is missing (as well as execution of it) on the new build slaves, resulting in ever growing docker images/layers exhausting all disk space quickly.
Please add the script and run it at least every 12, if not every 6 hours. We could run it from jenkins or via cron job.
Ideally, the script would be followed by "fstrim /" to make the SSDs perform faster. However, that requires sudo/root access.
On Tue, Apr 10, 2018 at 12:01:59PM +0000, lynxis [REDMINE] wrote:
Why not on the host?
much slower and more stress on the SSD. Basically, the discard all the time
is keeping your SSD busy with discarding those blocks (while you're trying
to compile something quickly to get a result for CI), and doing it periodically
in a cron job gives faster compile speeds. You can find various reports online
where people have been benchmarking.
Why is this still not working? We are again facing build failures due to out-of-space. It's really not acceptable that something like this (very small task / time requirement) is marked urgent and in the end is taking more than 10 days to resolve, sorry. We need to be able to rely on our infrastructure.