Re: [hlds_linux] [hlds] TF2 Replays

2014-05-11 Thread pilger
Apart from the FTP bug that cause the server to crash when it has any
problem shipping the replay files to the FTP server, the replays worked as
expected on my servers.

I'm running them on local http server mode and didn't have any problems
whatsoever.


_pilger


On 11 May 2014 09:58, Steven Sumichrast  wrote:

> I'm running replays on three servers 24/7 and haven't received complaints.
>  We've been using them since released and haven't turned them off besides
> for the FTP bug at the beginning with threads.
>
> That's maybe a difference -- we are using a FTP offload server. Are you?
>
> We clean up the FTP server nightly deleting replays over 14 days old.
>
> On Sunday, May 11, 2014, Erik-jan Riemers  wrote:
>
> > We dont have any issues either here, but we just have a crontab cleaning
> > old files at night when nobody is playing (makes more sense to me too)
> >
> >
> > 2014-05-11 12:30 GMT+02:00 Valentin G. 
> > >:
> >
> > > We've had no such issues and we are manually cleaning up the replay
> > > files with some find commands. We haven't changed that from way back
> > > when the auto cleanup was completely broken and I haven't heard of any
> > > similar complaints.
> > >  You may want to give that a shot.
> > >
> > > On Sun, May 11, 2014 at 12:16 PM, ics  wrote:
> > > > Ever since the guy who created replay on Valve moved on from TF2, the
> > > replay
> > > > system has been laid off and just kept as it is. There were some bugs
> > > left
> > > > in, such as replays not always getting deleted automatically and now
> > > > propably this where replays are deleted when noticed that they should
> > be
> > > but
> > > > removing them happens at bad time. This could explain the random lags
> > on
> > > > roud starts we've been having.
> > > >
> > > > I do clear up replay dirs occasionally manually but not recently. I
> > will
> > > > disable replay if this keeps up.
> > > >
> > > >
> > > > -ics
> > > >
> > > > Yun Huang Yong kirjoitti:
> > > >>
> > > >> Further info -- I observe that the two problems in the OP are
> > > correlated.
> > > >>
> > > >> During a recent pause I repeatedly looked at the replay directory
> and
> > > >> observe that the server deleted replays that it had just finished.
> > > >>
> > > >> Specifically:
> > > >>   - we were on cp_dustbowl
> > > >>   - there were multiple block files corresponding to the round just
> > > >> finished
> > > >>   - at the start of the next round, we had a 20 second pause
> > > >>   - during that 20 second pause, *all* of the block files for the
> > round
> > > >> just past were deleted
> > > >>
> > > >> We only keep 7 days worth of replays so it's a little baffling what
> > it's
> > > >> doing in any case -- stat()ing and unlinking ~10 files shouldn't
> take
> > 20
> > > >> seconds.
> > > >>
> > > >> I had these settings:
> > > >>
> > > >> replay_data_lifespan"7"
> > > >> replay_dofileserver_cleanup_on_start"1"
> > > >>
> > > >> Now commented out. Will have to wait for overnight restart of server
> > > >> (people are playing) to see it fixes things.
> > > >>
> > > >> Are there any other settings related to cleanup that I should
> disable?
> > > >>
> > > >> ics, do you have similar settings for cleanup?
> > > >>
> > > >> On 11/05/2014 6:34 PM, Erik-jan Riemers wrote:
> > > >>>
> > > >>> 15/20 seconds does not seem to sound like any "io" issues, since it
> > is
> > > >>> just
> > > >>> on round start, maybe its even some dns lookup that goes bad? (just
> > > >>> thinking out loud here) you could fix/found out by installing
> bind9,
> > > >>> point
> > > >>> your resolv.conf to it and placing it into debug mode to see the
> > logs.
> > > >>> (google for instructions, you can just let it forward to your
> regular
> > > dns
> > > >>> servers was like 5 minutes work) you never know!
> > > >>>
> > > >>> Also check your sysctl.conf if there are not strange "additions" to
> > it,
> > > >>> highly doubt there could be anything in there that would screw it
> > up...
> > > >>> but
> > > >>> again just throwing it out there.
> > > >>>
> > > >>>
> > > >>> 2014-05-11 9:36 GMT+02:00 Yun Huang Yong :
> > > >>>
> > >  Oh wow. For us it's happening across both of our 24p servers (we
> > only
> > >  have
> > >  2).
> > > 
> > >  And over the last few months it's gone from happening occasionally
> > > for a
> > >  few seconds, to happening on 80% of round changes, for 15-20
> > seconds.
> > > 
> > >  That suggests it may be related to something that might be growing
> > > with
> > >  time but I prune the replays directly externally so there's never
> > more
> > >  than
> > >  7 days worth of files.
> > > 
> > >  Do you run any of the following?
> > > - Sou
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
>
_

Re: [hlds_linux] [hlds] TF2 Replays

2014-05-11 Thread Erik-jan Riemers
i used to have ftp to offload, then i just created a http server on each
server since if the ftp is down, it could lock up your game if the
connection is lost for whatever reason.


2014-05-11 14:58 GMT+02:00 Steven Sumichrast :

> I'm running replays on three servers 24/7 and haven't received complaints.
>  We've been using them since released and haven't turned them off besides
> for the FTP bug at the beginning with threads.
>
> That's maybe a difference -- we are using a FTP offload server. Are you?
>
> We clean up the FTP server nightly deleting replays over 14 days old.
>
> On Sunday, May 11, 2014, Erik-jan Riemers  wrote:
>
> > We dont have any issues either here, but we just have a crontab cleaning
> > old files at night when nobody is playing (makes more sense to me too)
> >
> >
> > 2014-05-11 12:30 GMT+02:00 Valentin G. 
> > >:
> >
> > > We've had no such issues and we are manually cleaning up the replay
> > > files with some find commands. We haven't changed that from way back
> > > when the auto cleanup was completely broken and I haven't heard of any
> > > similar complaints.
> > >  You may want to give that a shot.
> > >
> > > On Sun, May 11, 2014 at 12:16 PM, ics  wrote:
> > > > Ever since the guy who created replay on Valve moved on from TF2, the
> > > replay
> > > > system has been laid off and just kept as it is. There were some bugs
> > > left
> > > > in, such as replays not always getting deleted automatically and now
> > > > propably this where replays are deleted when noticed that they should
> > be
> > > but
> > > > removing them happens at bad time. This could explain the random lags
> > on
> > > > roud starts we've been having.
> > > >
> > > > I do clear up replay dirs occasionally manually but not recently. I
> > will
> > > > disable replay if this keeps up.
> > > >
> > > >
> > > > -ics
> > > >
> > > > Yun Huang Yong kirjoitti:
> > > >>
> > > >> Further info -- I observe that the two problems in the OP are
> > > correlated.
> > > >>
> > > >> During a recent pause I repeatedly looked at the replay directory
> and
> > > >> observe that the server deleted replays that it had just finished.
> > > >>
> > > >> Specifically:
> > > >>   - we were on cp_dustbowl
> > > >>   - there were multiple block files corresponding to the round just
> > > >> finished
> > > >>   - at the start of the next round, we had a 20 second pause
> > > >>   - during that 20 second pause, *all* of the block files for the
> > round
> > > >> just past were deleted
> > > >>
> > > >> We only keep 7 days worth of replays so it's a little baffling what
> > it's
> > > >> doing in any case -- stat()ing and unlinking ~10 files shouldn't
> take
> > 20
> > > >> seconds.
> > > >>
> > > >> I had these settings:
> > > >>
> > > >> replay_data_lifespan"7"
> > > >> replay_dofileserver_cleanup_on_start"1"
> > > >>
> > > >> Now commented out. Will have to wait for overnight restart of server
> > > >> (people are playing) to see it fixes things.
> > > >>
> > > >> Are there any other settings related to cleanup that I should
> disable?
> > > >>
> > > >> ics, do you have similar settings for cleanup?
> > > >>
> > > >> On 11/05/2014 6:34 PM, Erik-jan Riemers wrote:
> > > >>>
> > > >>> 15/20 seconds does not seem to sound like any "io" issues, since it
> > is
> > > >>> just
> > > >>> on round start, maybe its even some dns lookup that goes bad? (just
> > > >>> thinking out loud here) you could fix/found out by installing
> bind9,
> > > >>> point
> > > >>> your resolv.conf to it and placing it into debug mode to see the
> > logs.
> > > >>> (google for instructions, you can just let it forward to your
> regular
> > > dns
> > > >>> servers was like 5 minutes work) you never know!
> > > >>>
> > > >>> Also check your sysctl.conf if there are not strange "additions" to
> > it,
> > > >>> highly doubt there could be anything in there that would screw it
> > up...
> > > >>> but
> > > >>> again just throwing it out there.
> > > >>>
> > > >>>
> > > >>> 2014-05-11 9:36 GMT+02:00 Yun Huang Yong :
> > > >>>
> > >  Oh wow. For us it's happening across both of our 24p servers (we
> > only
> > >  have
> > >  2).
> > > 
> > >  And over the last few months it's gone from happening occasionally
> > > for a
> > >  few seconds, to happening on 80% of round changes, for 15-20
> > seconds.
> > > 
> > >  That suggests it may be related to something that might be growing
> > > with
> > >  time but I prune the replays directly externally so there's never
> > more
> > >  than
> > >  7 days worth of files.
> > > 
> > >  Do you run any of the following?
> > > - Sou
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
>
___
To unsubscribe, edit your list preferences, or view the list archi

Re: [hlds_linux] [hlds] TF2 Replays

2014-05-11 Thread Steven Sumichrast
I'm running replays on three servers 24/7 and haven't received complaints.
 We've been using them since released and haven't turned them off besides
for the FTP bug at the beginning with threads.

That's maybe a difference -- we are using a FTP offload server. Are you?

We clean up the FTP server nightly deleting replays over 14 days old.

On Sunday, May 11, 2014, Erik-jan Riemers  wrote:

> We dont have any issues either here, but we just have a crontab cleaning
> old files at night when nobody is playing (makes more sense to me too)
>
>
> 2014-05-11 12:30 GMT+02:00 Valentin G. 
> >:
>
> > We've had no such issues and we are manually cleaning up the replay
> > files with some find commands. We haven't changed that from way back
> > when the auto cleanup was completely broken and I haven't heard of any
> > similar complaints.
> >  You may want to give that a shot.
> >
> > On Sun, May 11, 2014 at 12:16 PM, ics  wrote:
> > > Ever since the guy who created replay on Valve moved on from TF2, the
> > replay
> > > system has been laid off and just kept as it is. There were some bugs
> > left
> > > in, such as replays not always getting deleted automatically and now
> > > propably this where replays are deleted when noticed that they should
> be
> > but
> > > removing them happens at bad time. This could explain the random lags
> on
> > > roud starts we've been having.
> > >
> > > I do clear up replay dirs occasionally manually but not recently. I
> will
> > > disable replay if this keeps up.
> > >
> > >
> > > -ics
> > >
> > > Yun Huang Yong kirjoitti:
> > >>
> > >> Further info -- I observe that the two problems in the OP are
> > correlated.
> > >>
> > >> During a recent pause I repeatedly looked at the replay directory and
> > >> observe that the server deleted replays that it had just finished.
> > >>
> > >> Specifically:
> > >>   - we were on cp_dustbowl
> > >>   - there were multiple block files corresponding to the round just
> > >> finished
> > >>   - at the start of the next round, we had a 20 second pause
> > >>   - during that 20 second pause, *all* of the block files for the
> round
> > >> just past were deleted
> > >>
> > >> We only keep 7 days worth of replays so it's a little baffling what
> it's
> > >> doing in any case -- stat()ing and unlinking ~10 files shouldn't take
> 20
> > >> seconds.
> > >>
> > >> I had these settings:
> > >>
> > >> replay_data_lifespan"7"
> > >> replay_dofileserver_cleanup_on_start"1"
> > >>
> > >> Now commented out. Will have to wait for overnight restart of server
> > >> (people are playing) to see it fixes things.
> > >>
> > >> Are there any other settings related to cleanup that I should disable?
> > >>
> > >> ics, do you have similar settings for cleanup?
> > >>
> > >> On 11/05/2014 6:34 PM, Erik-jan Riemers wrote:
> > >>>
> > >>> 15/20 seconds does not seem to sound like any "io" issues, since it
> is
> > >>> just
> > >>> on round start, maybe its even some dns lookup that goes bad? (just
> > >>> thinking out loud here) you could fix/found out by installing bind9,
> > >>> point
> > >>> your resolv.conf to it and placing it into debug mode to see the
> logs.
> > >>> (google for instructions, you can just let it forward to your regular
> > dns
> > >>> servers was like 5 minutes work) you never know!
> > >>>
> > >>> Also check your sysctl.conf if there are not strange "additions" to
> it,
> > >>> highly doubt there could be anything in there that would screw it
> up...
> > >>> but
> > >>> again just throwing it out there.
> > >>>
> > >>>
> > >>> 2014-05-11 9:36 GMT+02:00 Yun Huang Yong :
> > >>>
> >  Oh wow. For us it's happening across both of our 24p servers (we
> only
> >  have
> >  2).
> > 
> >  And over the last few months it's gone from happening occasionally
> > for a
> >  few seconds, to happening on 80% of round changes, for 15-20
> seconds.
> > 
> >  That suggests it may be related to something that might be growing
> > with
> >  time but I prune the replays directly externally so there's never
> more
> >  than
> >  7 days worth of files.
> > 
> >  Do you run any of the following?
> > - Sou
___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux