sapinkumar amin created METRON-737:
--------------------------------------
Summary: Deploying additional model on an already running Maas
model creates ghost cotainers
Key: METRON-737
URL: https://issues.apache.org/jira/browse/METRON-737
Project: Metron
Issue Type: Bug
Affects Versions: 0.3.0
Environment: centos 6,7 .
java version "1.8.0_60"
Reporter: sapinkumar amin
Fix For: 0.3.0
Attachments: AppMaster.stderr.log, AppMaster.stderr.timeout.log,
container_successful_launch.log, mock_dga.zip
First deployment of Maas goes fine. If we ask for 1 container running model, it
does deploy total 2. One as AM and another as model itself. However adding more
number of containers of same model creates ghost containers in following fashion
1. $METRON_HOME/metron/bin/maas_deploy.sh -zq 'zookeeper_host:2181' -lmp
/root/mock_dga -hmp /user/root/models -mo ADD -m 512 -n dga1 -v 1.0 -ni 1
Creates 2 container. 1 is AM, another is model (mock_dga)
(Running same as above command)
2. $METRON_HOME/metron/bin/maas_deploy.sh -zq 'zookeeper_host:2181' -lmp
/root/mock_dga -hmp /user/root/models -mo ADD -m 512 -n dga1 -v 1.0 -ni 1
Creates 4 containers. 1 is AM, two are models (mock_dga). 1 is ghost.
(Running same as above command)
3. $METRON_HOME/metron/bin/maas_deploy.sh -zq 'zookeeper_host:2181' -lmp
/root/mock_dga -hmp /user/root/models -mo ADD -m 512 -n dga1 -v 1.0 -ni 1
Creates 7 containers. 1 is AM, 3 are models, 3 are ghosts.
Please find AppMaster stderr logs attached with this JIRA.
1. AppMaster.stderr.timeout.log has AM logs when after second attempts, one of
the ghost container times out. And total comes back to 3 which as expected.
2. AppMaster.stderr.log has AM logs about running command 3 times and launching
7 containers.
Please find the model (mock_dga.zip) and container launched log attached with
this jira. There are no logs for ghost container.
Please also note that if I issue REMOVE command, i.e.
$METRON_HOME/metron/bin/maas_deploy.sh -zq zookeeper_host:2181 -lmp
/root/mock_dga -hmp /user/root/models -mo REMOVE -m 512 -n dga1 -v 1.0 -ni 1,
It successfully removes the running model but not the ghost containers.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
