Would it be safe to trigger pending job scans on multiple machines parallel? We have performance issues related to this, and i was thinking maybe i could just set up 10-15 machines with cronjobs running this in the command line…
I wouldn’t recommend it. The process wouldn’t know of what the other processes are doing (ie, which jobs they’re checking), so you’d just have 10-15 full scans going on simultaneously, which would probably make the problem worse by performing redundant queries.
One thing you could try I guess, is moving this process off the pulse machine entirely by making another machine do the Dependency Check in a cron job every couple minutes, and just crank the Pulse check interval super high so that it rarely/never does it. Might get around any weird resource contention stuff Pulse might be running into…
Could this be parallelized in version7? This is becoming a bottleneck that we cant really solve…
To illustrate, single version iterations (one artist clicking one publish button) can spawn up to 130+ inter dependent simulation jobs. With 20+ sim artists, you can imagine that this becomes an issue very quickly :-\
It’s definitely something we can look at, we’re prioritizing bug-squashing for 7 right now, but we’ll probably go through another pass of feature requests once we get through those.
Either way, I’ve logged it internally 