Just to revive an old argument... I'm in the middle of a big cluster upgrade and reviewing the sysctl changes I deliberately make. In the process of this, I'm noting that Condor *lowers* a number of values that I suspect it thinks it is raising.
e.g. I setÂnet.core.rmem_max to 16MiB and condor lowers it to something very close to 10MiB (I think you screwed up a digit, it appears to be equal to 10*1014*1024).
Not that my applications are likely sensitive to this reduction but it's clear you're just blowing the settings out of the way even when they probably meet your needs. The following changes might help:
(1) Clearer decision making by Condor about whether to modify the setting
(2) Clearer .conf file in /etc/sysctl.d about which kernel parameters it's changing. You write names that you made up, not the actual kernel parameters.
In this particular example of net.core.rmem_max, I believe the action is not meaningful for TCP-based networks unless you also setÂnet.ipv4.tcp_rmem to a value that takes advantage of it. e.g. Â'4096 65536 16777216'.
![[Computer Systems Lab]](http://www.cs.wisc.edu/pics/csl_logo.gif){kind=logo}