When having lots of queues and lots of workers, the time it takes to serially start all workers gets noticeable. In my case this is >30s.
When trying to start them in parallel with e.g. cake cake_resque -q foo -n 5 & we run into troubles that sometimes a worker is reported as Fail during startup.
It's unlikely but not impossible that two workers receive the same value from microtime(true). The entropy is simply too low.
This $pidFile gets deleted after a worker successfully starts so although both workers start correctly up, one of them is faster deleting this file and the other is thus reported as being false (and: likely still to run correctly).
mfn
commented
2 years ago
When having lots of queues and lots of workers, the time it takes to serially start all workers gets noticeable. In my case this is >30s.
When trying to start them in parallel with e.g.
cake cake_resque -q foo -n 5 &we run into troubles that sometimes a worker is reported asFailduring startup.We traced this to the following code-line https://github.com/wa0x6e/Cake-Resque/blob/42c4fbcca3cd8e5ae7817604887e84208552f1d2/src/Shell/CakeResqueShell.php#L825 :
It's unlikely but not impossible that two workers receive the same value from
microtime(true). The entropy is simply too low.This
$pidFilegets deleted after a worker successfully starts so although both workers start correctly up, one of them is faster deleting this file and the other is thus reported as being false (and: likely still to run correctly).