Balrog Board owner 08/06/2018 (Mon) 13:21:05 No.9526 del
>>9525
I just asked Miao if we could try running with more RAM, since at this point I think we were jumping to conclusions and what's actually wrong is that our memcache is too small and our disk too slow. Notice how 500s aren't happening now; that's because without /sp/ to cause cache churn mongo isn't hitting the slow disk device so much that the whole system wedges waiting for I/O anymore.