Hi Trav, yes indeed your reference to "remembers" ..i.e. the STATE INFORMATION that has been mention many many times in this forum. is...
..... I guess for whatever reason, Compressor "Remembers" the failure, and won't write if it's accessing either the same failed drive or directory...
maintained in one of two places:
- ~/library/application support/compressor/storage/jobs/xxxxxxx-yyyyyyy (where xxxxxxx yyyyyyy is the current cluster ID. There may be several active and stale. We usually have two active.
sub directories Jobs Shared and sometimes hidden . files such as .TemporaryItems etc etc
- /Users/shared/library/application support/compressor/storage/xxxxxxxx-yyyyyyy (also where xxxxxxx yyyyyyy is the current cluster ID. There may be several active and stale)
not always yet sometime .hidden files to maintain some sharing handshake.. will stand corrected.
This is why when you brute force PURGE these structure, Qmaster/compressor is "reset' anew in a sense. You also need to review similar structures on other service nodes (other hosts) and perform the same ritual on these as well else this state information may persist and become quite bothersome.
As I mentioned earlier and in many posts, one needs to rm -rf these objects and rip them out from underneath compressor/qmaster structure as they are serialised when using the "empty trash" procedure. They won't always DELETE because they are in use. Qmaster will rebuilt these structures immediately.
Oh make sure you DISMOUNT (unmount/unshare) any of these file structures folders if they are specifically MOUNTed (network or direct via SAN) on other hosts (service nodes). As always Qmaster will recover.
Other posters have proposed a shutdown/restart (reboot) or terminate qmaster compressord subtastsks to release this serialisaion. Good idea, Yes this will work ofcourse yet is high disruptive.
I.M.O , For most Compressor.app V4.1 there's usually NO NEED to brute force/harshly (add your own superlative) DELETE the whole Compressor.app V4.1 (Dec 2013) ~/library/application support/compressor nor /Users/shared/library/application support/compressor structures to resolve your issues. It certainly works and used to work well in those unstable days of Qmaster. There's some much information by many experienced posters on these forums to reolve issues that will certainly return unless they can be readily resolved without a workaround. .
However it doesn't resolve the original issue and as we know can be disruptive if all the clusters (two in out=r case) are knocked out when in fact the jobs are running ok.
You can easily navigate the qmaster HISTORY structure in these folders and surgically remove bothersome jobs.
HTH
Post your results for others to see.
warwick
Hong Kong