HQueue hquser crashing

[ptunstall]

Hey,

Trying to figure out why octane crashes when launched via hqueue, hython or hscript as the "hquser" username.

Code: Select all

[hquser@weeWhammy hfs18.0]$ cd /opt/hfs18.0 && source houdini_setup
The Houdini 18.0.597 environment has been initialized.
[hquser@weeWhammy hfs18.0]$ hscript
[Octane] Octane Render for Houdini. Build 2020.1.5.3 (Build date: Nov  5 2020 20:37:21)
[Octane] Octane API: 2020.1.5
[Octane] HDK API: 18.0.597
[Octane] Houdini host version: 18.0.597
[Octane] Initializing Octane API
14938 ThreadId=0x7fa6aebff700: Fatal error: Segmentation fault
-- TRACEBACK BEGIN --
Traceback from hbatch 18.0.597 (Compiled on linux-x86_64-gcc6.3):
stackTrace() <libHoudiniUT.so>
signalCallback(UTsignalHandlerArg) <libHoudiniUT.so>
UT_Signal::UT_ComboSignalHandler::operator()(int, siginfo_t*, void*) const <libHoudiniUT.so>
UT_Signal::processSignal(int, siginfo_t*, void*) <libHoudiniUT.so>
_L_unlock_13 <libpthread.so.0>
Octane::maxHosts(Octane::NetInterfaceInfo const&) <liboctane.so>
Octane::maxHosts(Octane::NetInterfaceInfo const&) <liboctane.so>
Octane::maxHosts(Octane::NetInterfaceInfo const&) <liboctane.so>
Octane::maxHosts(Octane::NetInterfaceInfo const&) <liboctane.so>
Octane::maxHosts(Octane::NetInterfaceInfo const&) <liboctane.so>
Octane::maxHosts(Octane::NetInterfaceInfo const&) <liboctane.so>
Octane::runFeaturesCheck() <liboctane.so>
Octane::runFeaturesCheck() <liboctane.so>
Octane::runFeaturesCheck() <liboctane.so>
Octane::ApiProjectWorkspace::getNodeGraphEditors() <liboctane.so>
Octane::ApiFileChooser::results() const <liboctane.so>
start_thread <libpthread.so.0>
__clone <libc.so.6>
-- TRACEBACK END --
14938: Fatal error: Segmentation fault (sent by pid 14938)
-- TRACEBACK BEGIN --
Traceback from hbatch 18.0.597 (Compiled on linux-x86_64-gcc6.3):
stackTrace() <libHoudiniUT.so>
signalCallback(UTsignalHandlerArg) <libHoudiniUT.so>
UT_Signal::UT_ComboSignalHandler::operator()(int, siginfo_t*, void*) const <libHoudiniUT.so>
UT_Signal::processSignal(int, siginfo_t*, void*) <libHoudiniUT.so>
_L_unlock_13 <libpthread.so.0>
__pthread_cond_wait <libpthread.so.0>
Octane::apiMode_Shared_start(char const*, bool, char const**, char* (*)(void*), void*, bool) <liboctane.so>
HOctane_initAPI(bool) <Houdini_Octane_18.0.597.so>
ROP_Octane::getTemplates() <Houdini_Octane_18.0.597.so>
ROP_Octane::getTemplatePair() <Houdini_Octane_18.0.597.so>
newDriverOperator <Houdini_Octane_18.0.597.so>
UT_DSO::run(char const*, void*, bool) <libHoudiniUT.so>
OP_OperatorTable::runDSOInstall() <libHoudiniPRM.so>
OP_Network::getOperatorTable(UT_StringRef const&, char const*) <libHoudiniPRM.so>
MGR_Node::buildRopTable(char const*) <libHoudiniOPZ.so>
MGR_Node::initTable(MGR_MANAGER_TYPE) <libHoudiniOPZ.so>
MOT_Director::MOT_Director(UT_StringHolder const&, bool, bool) <libHoudiniOPZ.so>
_init <hbatch-bin>
__libc_start_main <libc.so.6>
_init <hbatch-bin>
-- TRACEBACK END --

when launched as the main user, "ptunstall", I don't have this issue when rendering from both hython and hscript.

I've added hquser to the "wheel" group as this is the way to give users SUDO privilege's in centos.

I've also chmod -R 777 /opt/Octane* so there shouldn't be any permission issues with octanes install path.

Do you know why this might be happening?

[juanjgon]

Hmm, nope, sorry, I have no idea about what could be the problem. Do you know if the plugin running inside Houdini or even if Standalone works fine using this new user?

Thanks,
-Juanjo

[ptunstall]

These are headless nodes. Maybe it has to do with that cuda download thing. Where does it save that file on linux?

[juanjgon]

I'm not sure, to be honest. I've installed the latest Standalone on Linux and it works without downloading anything, so I wonder if this file is part of the nvidia drivers on Linux, or if I already had it on my system. Have you ever seen Standalone download it on Linux? In this case, the final install directory should be printed in the download window.

Thanks,
-Juanjo

[ptunstall]

Appreciate the time. I'm going to start with a fresh centos build and see what happens and I'll share my findings here.