Re: [HTCondor-users] Reducing the 12 second sleep?

[Jim White <jimwhite@xxxxxx>]

This is big peeve of mine as well. A solution would be really nice. I've not looked into what exactly the issue being addressed is here but there are a many distributed uuid algorithms and implementations that don't need a delay.

Jim

On Mon, Dec 8, 2014 at 10:23 AM, Brian Candler <b.candler@xxxxxxxxx> wrote:

...
12/08/14 16:53:26 Sleeping for 12 seconds to ensure ProcessId uniqueness
12/08/14 16:53:38 Warning: ProcessId not confirmed unique
...

Does anyone know how to reduce this? It seems arbitarily long and is a problem when using SUBDAG EXTERNAL, as each subdag also adds another 12 seconds of delay.

I found ProcessId::computeWaitTime and I see that the wait is calculated as
(3.0 * precision_range / time_units_in_sec)

I also found ProcAPI::createProcessId, which takes 4 arguments, the last of which is int* precision_range, but util_create_lock_file only calls it with 3. So it uses the defaults:

int ProcAPI::DEFAULT_PRECISION_RANGE = 4;

#if defined(HZ)
double ProcAPI::TIME_UNITS_PER_SEC = (double) HZ;
#elif defined(_SC_CLK_TCK)
double ProcAPI::TIME_UNITS_PER_SEC = (double) sysconf(_SC_CLK_TCK);
#else // not linux
double ProcAPI::TIME_UNITS_PER_SEC = 1.0;
#endif

Therefore I guess TIME_UNITS_PER_SEC = 1.0 on both systems I'm using (Linux and OSX).

Is the time resolution really that poor? And is this just to get a unique condor pid?

Regards,

Brian Candler.

_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@cs.wisc.edu with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/htcondor-users/