thread pools will flush their cached (dead) threads after 60 seconds. This will cause that YaCy now runs constantly withl about 50 threads, about 100 at peak times. Previously, about 400 threads had been cached and kept in a hibernation state, which caused that the numproc counter in /proc/user_beancounters (exists only in VM-hosted linux) was as high as the cached number of threads. This caused that VM supervisors terminated whole VM sessions if a limit was reached. Many VM providers have limits of numproc=96 which made it virtually impossible to run YaCy on such machines. With this change, it will be possible to run many YaCy instances even on VM hosts.pull/1/head
parent
181911376c
commit
321840fde3
Loading…
Reference in new issue