As you may know, we're having storage issues with the 8GB default Node disk (also see SCALE-23); a very high frequency of requests coming into the nodes leads access logs to grow extremely fast.

While checking through the settings in /opt/atlassian/bitbucket/current/app/WEB-INF/classes/logback.xml it seems to me the defaults do not facilitate this and would lead to a node filling up and failing. As default settings seems to 30 days, 10 files of 25 MB each day means there is nowhere near sufficient space available so by default there is a requirement to tweak these config files. 

Although tuning thelogback.xml  settings is possible, these tweaks disappear as this file is transient in nature (= upon instance termination it goes back to default on a newly spun node).

I'd like to propose that storing log- (and maybe other settings?) related configuration to be stored on the SHARED fileserver so that settings persist through node terminiations/recreations.

Note: This is somewhat related to SCALE-22 actually as this also requests to make nodes more ''termination resistant' ...