On 10 Mar 2019, Adam Goryachev said: > "My array freezes for 2 or 3 minutes every couple of weeks, usually I > don't notice, but its a little annoying when I do. Anyone know why" You probably would notice. A lot of the time the freeze would happen when you were reading, which means whatever you were waiting for, you're now going to wait for it for minutes. Any other reads from the same array will pile up behind that one, and if that includes binaries you're quite likely to end up with an apparently-frozen machine, kernel hangcheck timer warnings, etc. (Heck, possibly even actual automatic reboots from watchdog timer firings if that is one of the things that ends up blocked trying to read a pidfile or something.) When this happened to me on my desktop in times gone by an apparently-frozen X server was common (it was blocked on core fonts reads or paging or some X client that blocked in the middle of what was *meant* to be an infinitesimally short server grab... and that failure mode means that an X client on a remote server with RAID can trigger this on your desktop if you're unlucky enough.) Local filesystem storage does not normally -EINTR and is meant to respond "effectively immediately". When it doesn't, bad things happen. (Even a 7sec timeout is enough to be seriously noticeable.) -- NULL && (void)
