Hserver Communications Error

   16968   17   1
User Avatar
Member
48 posts
Joined: 7月 2005
Offline
I just upgrade three PCs to H7.0.341 and tried running a render test using all three (specified as Hosts on the Mantra command line).

Hscript and the license server are on the same PC (win2K SP2) (others are Win XP)

All seemed well at first. After about 15 minutes the render stopped with the message:

License: Hserver communication error

This is strange since the license server and hscript are on the same PC.

Then when I tried to open a .hip file, nothing happens. I see that there's an hserver.exe and a mantra.exe listed in the process list (although they are not changing size). So I'm guessing that hserver died.

Has anyone else encountered this? What can I do?
Adtech Communications Group
User Avatar
スタッフ
269 posts
Joined: 7月 2005
Offline
This sounds like the bug that existed in versions before 7.0.341 but was supposed to be fixed in 7.0.341. Does hserver die every time you do this render combination? If you go back to an older version, does the problem exist there?
User Avatar
Member
48 posts
Joined: 7月 2005
Offline
I upgraded the license server PC (Win2K) to Service Pack 4. Now I don?ft get the ?gcommunications error?h, but it does still hang from time to time when rendering with more than one PC (specified in mantra?fs hosts command). When I go into services and Stop/Start Houdini License Server then the renders come alive again.

I used this exact combination of PCs with H6x (many, many times) and never had this problem.

It does die everytime with H7 License server, although how many frames it takes to die varies.

I also did a test running a render on two PCs (one was the license server), when I burned a CD on the license server PC, the renders hung. Although it was a bit of an extreme test since I won't be burning CDs during renders it showed that it isn't only license serving that kills it …. other intensive processes running simultaneously can kill it too.

Is mantra waiting for data from hserver and waits indefinitely until it receives it? The question is why isn?ft hserver sending the data and why doesn't hserver try several times?

Any more ideas? I really need to get this fixed.
Thanks
Adtech Communications Group
User Avatar
Member
48 posts
Joined: 7月 2005
Offline
An additional note: When i look at the service called “Houdini License Server” I see that the actual exacutable is “sisinitd.exe” … and this is what I STOP/START to get the renders moving again.

Oddly, the service called “Houdini License Client”'s exacutable is “hserver.exe”.
Adtech Communications Group
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
Has anyone got a fix for this error
Im running a dual xeon system which is the
license server as well.
when I make mantra render with both procs
(-H localhost,localhost)
once in a while during a batch render I get

Warning: License server communication failure. Termination in 3261 seconds

and several frames dont render
then it fixes itself and continues rendering
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
Has anyone got a fix for this error
Im running a dual xeon system which is the
license server as well.
when I make mantra render with both procs
(-H localhost,localhost)
once in a while during a batch render I get

Warning: License server communication failure. Termination in 3261 seconds

and several frames dont render
then it fixes itself and continues rendering
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
Has anyone got a fix for this error
Im running a dual xeon system which is the
license server as well.
when I make mantra render with both procs
(-H localhost,localhost)
once in a while during a batch render I get

Warning: License server communication failure. Termination in 3261 seconds

and several frames dont render
then it fixes itself and continues rendering
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
Has anyone got a fix for this error
Im running a dual xeon system which is the
license server as well.
when I make mantra render with both procs
(-H localhost,localhost)
once in a while during a batch render I get

Warning: License server communication failure. Termination in 3261 seconds

and several frames dont render
then it fixes itself and continues rendering
User Avatar
Member
7740 posts
Joined: 7月 2005
Offline
Which version of Houdini are you using? How many render licenses do you have?
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
Im using houdini 7 and I have 2 render licenses
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
I have to assume that with both processors busy
rendering that a license request is timing out.

Im seeing three mantra process per render
I assume 1 mantra process spawns the two multiproc
mantra render process

In an attempt to fix this problem Im nice'ing the hscript batch script command to priority 1
(nice +1 DOscript.sh ….)
This sets hscript and 1 mantra priority to 1
but the two mantra process that do the actual render do not inherit the priority level
and stay at a higher priority of 0.

Does mantra keep checking for licenses after it
starts rendering. If it does than a exit code check
would not fix this issue would it?
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
I spotted this forum email on ofForce(see below)
(its for windows Houdini 6 but I see the same error
in my error log)
no slution there either

odforce
http://odforce.net/forum/index.php?showtopic=2133&pid=14374&mode=threaded&show=&st=&#entry14374 [odforce.net]
Oct 21 2004, 11:51 AM
Post #6
UT_NetPacket::write: Invalid argument
mantra: Unable to write data in network handshake

This is a common error I get, when I try to render with the “-H” flag.

I'm running (on both machines):
Win2000-SP4 (with latest patches)
Houdini 6.1.208 (no other version Houdini services running)
Calling the same drive/location ($HIP = “P:\Misc\Tests”)

Could anyone help me? Thanks in advance!
User Avatar
Member
7740 posts
Joined: 7月 2005
Offline
If you look at lucas post about this thread, this problem was fixed in Houdini 7.0.341.
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
Im using houdini 7.0.426
which part of this issue was fixed?
User Avatar
Member
7740 posts
Joined: 7月 2005
Offline
What RobertP said in his first post which sounds like the same problem as yours.
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
Sorry i was being facetious … and not very funny
Im using a houdini version later than the one Lucas
mentioned and Im still having the error
User Avatar
Member
4140 posts
Joined: 7月 2005
Offline
Hey Will - yah I would agree with your guess that something is overlapping in the process…have you verified things like memory/pagefile usage to see if this is related to other system problems causing timeouts while swapping, for example? (not that I have any reason to believe you'd need lots of mem… ). In other words, can you render a long animation of a plastic ball and still have this happen occasionally, or does it only happen during high stress? At least that can whittle it down to being an inherent problem with mantra not shutting down properly as opposed to not behaving nicely with licenses in stress situations.

Cheers,

J.C.
John Coldrick
User Avatar
Member
98 posts
Joined: 7月 2005
Offline
Excellent diagnosis Dr. Coldrick.
As long as nothing gets into swap I can dual proc
render.
SO I guess the badness of swap hell gets the
the license program all flustered.
Since rendering ram spikes can occur even in the most miserly scenes … I would still think this happens
often enough to warrant a more robust solution.
In this case I pre-created IFD files and mantra rendered
without an hscript running which cleared up a lot of ram.

thanks
  • Quick Links