[
https://issues.apache.org/jira/browse/MESOS-2323?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14379960#comment-14379960
]
Till Toenshoff commented on MESOS-2323:
---------------------------------------
After giving a "ship-it" on those reviews, I am having second thoughts after
looking at the output - here is a simple example:
{noformat}
I0325 15:08:06.146759 194703360 master.cpp:358] Master
20150325-150806-347252928-5050-32055 (lobomacpro2.fritz.box) started on
192.168.178.20:5050
I0325 15:08:06.146787 194703360 master.cpp:360] Flags at startup:
{"allocation_interval":"1secs","authenticate":"false","authenticate_slaves":"false","authenticators":"crammd5","framework_sorter":"drf","help":"false","initialize_driver_logging":"true","log_auto_initialize":"true","logbufsecs":"0","logging_level":"INFO","port":"5050","quiet":"false","recovery_slave_removal_limit":"100%","registry":"replicated_log","registry_fetch_timeout":"1mins","registry_store_timeout":"5secs","registry_strict":"false","root_submissions":"true","slave_reregister_timeout":"10mins","user_sorter":"drf","version":"false","webui_dir":"/Users/till/Development/mesos/build/../src/webui","work_dir":"/tmp/mesos","zk_session_timeout":"10secs"}
{noformat}
While this certainly contains all needed information for debugging, it is
definitely not nice to read especially considering that it is wrapped multiple
times in a console. Our logs are meant to help debugging with human readable
output. The above IMHO is rather noisy and hard to read without a JSON parser.
I would love to see one value-pair per line. That however would certainly add a
lot of log-lines for this feature.
It is really hard for me to give a definite answer. Right now I feel that
adding another flag to master & slave triggering a nicely formatted flag setup
dump would be nicer. So those wo want/need such feature can enable it.
{{mesos-master.sh --show_flags ...}}
{noformat}
Flags at startup:
--allocation_interval="1secs"
--authenticate=false
--authenticate_slaves=false
--authenticators="crammd5"
--framework_sorter="drf"
--help=false
--initialize_driver_logging=true
--log_auto_initialize=true
--logbufsecs=0
--logging_level=INFO
--port=5050
--quiet=false
--recovery_slave_removal_limit="100%"
--registry="replicated_log"
--registry_fetch_timeout="1mins"
--registry_store_timeout="5secs",
--registry_strict=false
--root_submissions=true
--slave_reregister_timeout="10mins"
--user_sorter="drf"
--version=false
--webui_dir="/Users/till/Development/mesos/build/../src/webui"
--work_dir="/tmp/mesos"
--zk_session_timeout="10secs"
{noformat}
What do you think?
> write flags to log at startup
> -----------------------------
>
> Key: MESOS-2323
> URL: https://issues.apache.org/jira/browse/MESOS-2323
> Project: Mesos
> Issue Type: Improvement
> Reporter: David Robinson
> Assignee: Joerg Schad
> Priority: Minor
>
> Whilst troubleshooting problems it's sometimes useful to see the exact flags
> that were set when Mesos started. The flags are exposed via the /state.json
> endpoint whilst Mesos is running, but it's not possible to know exactly what
> flags were set during a previous run. Writing this information to the logs
> during startup would make troubleshooting problems much easier.
> Some of the flag's settings can be inferred from log entries, but some, such
> as --attributes can not.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
