1500 is our user base, not the OP's - just want to make that clear. We're doing
just fine. We've not seen any problems with the polling from Perforce clients; It's
just not a big impact item. (And we've been at this level for about nine months,
plenty of time for a large cluster of P4Vs to hit at once - I'm sure they've done
so many times.) If there have been occasional slowdowns due to polling, they've
been too transient to really register.
I should mention that our 1500 users are spread over about 13 server instances
on a single linux system. (Not ideal, but a result of numerous migrations from
an assortment of SCM systems and unavoidable at this point.)
So it is actually handling a much higher over-all load than most hardware, yet, the
Perforce side of it works just fine and handles the load with plenty of room to
spare. (It is a killer piece of hardware.) Our only major impacts came from a bad
network connection (which caused p4d processes to queue up because they
couldn't communicate efficiently) and the poor performance of the one trigger I
mentioned. Even the script that was banging away 8-12 times a second wasn't
noticed until the network issues brought it to the surface.
-Wm
From: ***@gmail.com [mailto:***@gmail.com] On Behalf Of Jeff A. Bowles
Sent: Thursday, March 11, 2010 5:39 PM
To: Ivey, William
Cc: Rick Macdonald; perforce-***@perforce.com
Subject: Re: [p4] perforce client process
There's a piece of information missing.
If someone's running an ill-configured Hudson server against the Perforce server, for example, there might be expensive queries beating on the server. (I'd bet on something like that.)
(If it's 1500 people hitting refresh on p4win/p4v at the same time, well, let's think about junior high band camp: everyone flushes the toilets in the dorms at the same time, and the plumbing breaks in the building when they do so.... This is similar.)
-Jeff Bowles
On Thu, Mar 11, 2010 at 2:04 PM, Ivey, William <***@bmc.com<mailto:***@bmc.com>> wrote:
We have 1500 users and I don't think the polling has been much
of an issue as far as performance goes. We did have a runaway
script of some sort that was pinging the server several times a
second, but even that wasn't killing it except that it happened
at a time when there were network issues creating an additional
bottleneck. Diagnosing that issue turned up the heavy pinging.
Right now our biggest problem seems to be a trigger designed to
interface to Rally. It appears to hold up submits until it has
performed several transactions with an external server which
seems to take a lot longer than it should. Response after a
submit has gone from 1-2 seconds to 10-20 seconds. That's masking
all other possible performance hits, if there are any. So if I
was looking for performance issues, I'd check triggers first.
-Wm
-----Original Message-----
From: perforce-user-***@perforce.com<mailto:perforce-user-***@perforce.com> [mailto:perforce-user-***@perforce.com<mailto:perforce-user-***@perforce.com>] On Behalf Of Rick Macdonald
Sent: Thursday, March 11, 2010 12:47 PM
To: perforce-***@perforce.com<mailto:perforce-***@perforce.com>
Subject: Re: [p4] perforce client process
We have 270 licenses. Not everybody runs p4v/p4win all day, but it was
suspected that all the open apps doing the automatic refresh every 5
minutes was a noticeable load on the server. Each refresh of course is
several queries to the sever. We have a wrapper script for p4v that
installs my custom tool menu, so I have this script set the refresh to 0
in the user's settings file to turn of the refresh.
We just went ahead and did this without any specific measurements before
and after. Does anybody have a feel for how much of a load this can be?
Rick
Post by Matt JanulewiczI'd also like to add that if your user processes are greatly affecting
the performance of the server, it's time to spend some money on hardware.
As others have indicated, connections to the server (or client
processes) are established at the onset of a command and terminated
when it's done. Perforce client apps do not hold connections open. So
it would have to be a pretty unique situation where many users were
doing many 'heavy' queries at the same time for people to notice.
Or, perhaps, you have 100 people pointed at a 386 or virtual machine
that has no resources to begin with.
-Matt Janulewicz
Lucasfilm Entertainment Company Ltd.
Post by saurabh talwarHi Prashant
You need to be clear as to which client process you require
"stopping". Are
you using P4V?
I wonder if you are referring to the fact that P4V by default is set up to
poll the server regularly (typically every 5 minutes) for latest updates.
If you have lots of users doing this it starts to place a load on the
server.
It is easy to turn off, put 0 in the appropriate field: Edit>
Preferences>
Connection> "Check Server for updates"
It is also possible to create an auto-config to do this as part of the
installer. Can give you a pointer if that is relevant.
Robert
Post by Ivey, William-----Original Message-----
Sent: 11 March 2010 04:45
Subject: [p4] perforce client process
Hi,
How can I stop and restart my client process in perforce ?
Thanks in Advance,
Prashant H
_______________________________________________
http://maillist.perforce.com/mailman/listinfo/perforce-user
_______________________________________________
http://maillist.perforce.com/mailman/listinfo/perforce-user
_______________________________________________
http://maillist.perforce.com/mailman/listinfo/perforce-user
_______________________________________________
perforce-user mailing list - perforce-***@perforce.com<mailto:perforce-***@perforce.com>
http://maillist.perforce.com/mailman/listinfo/perforce-user
_______________________________________________
perforce-user mailing list - perforce-***@perforce.com<mailto:perforce-***@perforce.com>
http://maillist.perforce.com/mailman/listinfo/perforce-user