Jump to content

Miles O'Neal

Member Since 28 Oct 2014
Online Last Active Today, 08:44 PM

Topics I've Started

monitor: p4 configure vs p4 counter

18 September 2020 - 11:00 PM

It appears that "p4 configure set monitor=1" and "p4 counter -f monitor 1" affect the same piece of server data. Is that correct? If so, why?

I am using 2020.1 server and client.


single ticket login (auth.id) question

31 August 2020 - 08:29 PM

Just to clarify, if we start using the auth.id-based single ticket login, login checks are not sent back to the master, correct? As soon as the master propagates the ticket to a replica (based on rpl.forward.login being set), login checks for that replica happen right there?


Retrieve number of files sync'd to a client

24 July 2020 - 05:25 PM

Is there a more efficient way to get the number of files currently sync'd to a client or opened on a client? Right now I'm doing

p4 -c <client> have //<client>/... | wc -l
p4 opened -C <client> | wc -l

I can't find a command to just return a count rather than the list of files.


To hyperthread or not?

09 July 2020 - 10:38 PM

The only reference I could find to hyperthreading and Helix was a vague, minimalist KB article:

Our performance lab has tested the Perforce server with hyper-threading. The results indicate that hyper-threading does not improve Perforce server performance. While there is no evidence that hyper-threading has any negative impacts on performance, we recommend leaving hyper-threading disabled.

What, exactly, was tested? What were the results? Was the number of jobs ramped up past the number of physical cores with and without hyperthreading enabled, and no difference in results?

If there are no negative impacts, why leave it off?

Some actual results would be really nice. I realize that MMMV.

Has anyone here been involved in stress-testing p4d running on a modern (last year or so) server running RHEL7 (or any 3.x Linux), with and without hyperthreading? If so, what did you find?

We've been running hefty Intel-based servers with hyperthreading on the past 6-7 years with no obvious ill effects.


obliterate would only destroy 35 out of 9905 files?

18 June 2020 - 10:42 PM

I'm trying to test obliterate times on an older p4d vs a new one. I've cloned the production data and was ready, I thought, for the timed tests. But "p4 obliterate" (no "-y" yet) thinks it only needs to destroy 35 files (including versions), whereas "p4 files" tells me there are 9905 files there. Only 31 of the files are not "branch change" files; these seem to be the ones being obliterated. Obliterate with "-b" lists no files.

"p4 snap -n" found no files that need to be snapped.

As expected, it makes no difference in the output whether I use a client or not. (Client is faster, of course,)

% p4 obliterate //moneal_reusable_oblit/...
[35 files listed]
Would delete 1818296 client 19760 integration 9935 revision record(s).
This was report mode.  Use -y to remove files.
% p4 obliterate -b //moneal_reusable_oblit/...
Would delete 1816816 client 19748 integration 9874 revision record(s).
This was report mode.  Use -y to remove files.

What's going on here? Is it really going to delete all the files but just not list them?