[
https://issues.apache.org/jira/browse/SINGA-150?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15209756#comment-15209756
]
Chen Peng commented on SINGA-150:
---------------------------------
I create a fork from the official repo and comment out the
`InitGoogleLogging()` method in `singa_scheduler.cc`. When training in
pseudo-distributed mode (4 nodes in total, node0 to node3), the program hangs
with the following output
root@node0:~/incubator-singa/tool/mesos# ./scheduler
/root/incubator-singa/examples/cifar10/job.conf -
scheduler_conf ./scheduler.conf -singa_conf
/root/incubator-singa/conf/singa.conf
WARNING: Logging before InitGoogleLogging() is written to STDERR
I0323 21:53:18.856851 277 singa_scheduler.cc:413] Scheduler initialized
I0323 21:53:18.869976 277 singa_scheduler.cc:419] Starting SINGA framework...
I0323 21:53:18.878701 277 sched.cpp:157] Version: 0.22.0
I0323 21:53:18.888286 282 sched.cpp:254] New master detected at
[email protected]:5050
I0323 21:53:18.891774 282 sched.cpp:264] No credentials provided. Attempting
to register without aut
hentication
I0323 21:53:18.894948 283 sched.cpp:448] Framework registered with
20160323-214924-33558956-5050-144
-0000
I0323 21:53:18.896631 288 singa_scheduler.cc:356] n1 = 1 n2 = 1 ncpus = 2
I0323 21:53:18.896728 288 singa_scheduler.cc:356] n1 = 1 n2 = 1 ncpus = 2
I0323 21:53:18.896771 288 singa_scheduler.cc:356] n1 = 1 n2 = 1 ncpus = 2
The log in slave node is not human readable as
~U^@^@^@
^Enode3^Z^V
^Dcpus^P^@^Z ^@^@^@^@^@^@^@@2^A*^Z^U
^Cmem^P^@^Z ^@^@^@^@^@ަ@2^A*^Z^V
^Ddisk^P^@^Z ^@^@^@^@^@e2^A*^Z^X
^Eports^P^A"
^H^H~X[34m^A^P~@^A2^A*2&
$20160323-214924-33558956-5050-144-S28^A@�'
> "called InitGoogleLogging() twice!" error
> -----------------------------------------
>
> Key: SINGA-150
> URL: https://issues.apache.org/jira/browse/SINGA-150
> Project: Singa
> Issue Type: Bug
> Environment: Ubuntu 12.04.5 LTS (GNU/Linux 3.13.0-83-generic x86_64)
> Reporter: Chen Peng
> Assignee: Anh Dinh
>
> When launching SINGA on pseudo-distributed mode (single node) according to
> http://singa.apache.org/docs/docker.html and train it according to
> http://singa.apache.org/docs/mesos.html, two nodes (node0 and node1) are
> created.
> When executing the command `./scheduler /root/incubator-singa/│install.sh
> zookeeper-3.4.6 zookeeper-3.4.6.tar.gz
> examples/cifar10/job.conf -scheduler_conf ./scheduler.conf -singa_conf
> /roo│root@node1:~/incubator-singa/thirdparty# uname
>
> t/incubator-singa/conf/singa.conf ', the following error is displayed:
> I0317 23:08:19.499140 298 singa_scheduler.cc:413] Scheduler initialized
> F0317 23:08:19.503722 298 utilities.cc:317] Check failed: !IsGoogleLoggin
> gInitialized() You called InitGoogleLogging() twice!
> *** Check failure stack trace: ***
> @ 0x7f9178664612 google::LogMessage::Fail()
> @ 0x7f917866455e google::LogMessage::SendToLog()
> @ 0x7f9178663f60 google::LogMessage::Flush()
> @ 0x7f9178666e74 google::LogMessageFatal::~LogMessageFatal()
> @ 0x7f917866ac21 google::glog_internal_namespace_::InitGoogleLoggi
> ngUtilities()
> @ 0x7f917866717f google::InitGoogleLogging()
> @ 0x7f9177ccff6d mesos::internal::logging::initialize()
> @ 0x7f9177eba176 mesos::MesosSchedulerDriver::initialize()
> @ 0x7f9177eba95c mesos::MesosSchedulerDriver::MesosSchedulerDriver
> ()
> @ 0x408c3d main
> @ 0x7f9174109ec5 (unknown)
> @ 0x408669 (unknown)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
