You are browsing a read-only backup copy of Wikitech. The primary site can be found at

Analytics/Systems/Cluster/Camus: Difference between revisions

From Wikitech-static
Jump to navigation Jump to search
(Reflect that camus runs on an-launcher1002)
imported>Neil P. Quinn-WMF
(Neil P. Quinn-WMF moved page Analytics/Systems/Cluster/Camus to Obsolete:Camus in the Analytics Cluster: Camus has been replaced by Gobblin)
Line 1: Line 1:
This info is for members of analytics team.
#REDIRECT [[Obsolete:Camus in the Analytics Cluster]]
Analytics Production Camus jobs are launched via hdfs user cron on an-launcher1002 (check site.pp in Puppet first for the role analytics_cluster::launcher).
As of September 2020, possible replacements for Camus are being evaluated ([[phab:T238400|T238400]]).
== How to stop Camus ==
The quickest way is to ssh to an-launcher1002 and check/stop systemd timers:
ssh an-launcher1002.eqiad.wmnet
sudo systemctl list-timers | grep camus
Then disable puppet (requires SRE/root permissions) and:
sudo systemctl stop camus-webrequest
== Check Camus Production logs ==
* ssh to an-launcher1002
* logs are stored in /var/log/camus, one (rotated) file per camus run-type (as of today: <code>webrequest</code>, <code>eventlogging</code>, <code>mediawiki</code> and <code>eventbus</code>)
* In those files are logged both camus output and camus-partition-checker output.
== How to produce to kafka ==
cat test_message.txt  | kafkacat -b  kafka-jumbo1001.eqiad.wmnet:9092 -t test
Test message is a file like:
{"id":123456,"name":"pepito perez", "muchoStuff":{"a": "1"}}
{"id":123456,"name":"pepito perez", "muchoStuff":{"a": "2"}}
{"id":123456,"name":"pepito perez", "muchoStuff":{"a": "3"}}
{"id":123456,"name":"pepito perez", "muchoStuff":{"a": "4"}}
== How to validate your data against your avro schema ==
We have found php bindings to be different than java ones, please validate messages using this java jar:
java -jar avro-tools-1.7.6.jar jsontofrag --schema-file CirrusSearchRequestSet.avsc searchmessage.json
== How to run camus job to decode avro from kafka topic ==
Camus is our map reduce job but also has some of the code we depend on, thus camus jar appears twice.
Note that you need your local properties file to pass to camus. Note: "-P /home/user/avro-kafka/" below
"Real" properties files live on puppet: []
export LIBJARS=/home/user/avro-kafka/camus-wmf-0.1.0-wmf6.jar,/home/user/avro-kafka/camus-etl-kafka-0.1.0-wmf6.jar,/home/user/avro-kafka/camus-api-0.1.0-wmf6.jar,/home/user/av
export HADOOP_CLASSPATH=/home/user/avro-kafka/camus-wmf-0.1.0-wmf6.jar:/home/user/avro-kafka/camus-etl-kafka-0.1.0-wmf6.jar:/home/user/avro-kafka/camus-api-0.1.0-wmf6.jar:/home/user/avro-kafka/camus-kafka-coders-0.1.0-wmf6.jar:/home/user/avro-kafka/camus-schema-registry-0.1.0-wmf6.jar:/home/user/avro-kafka/camus-parent-0.1.0-wmf6-tests.jar:/home/user/avr
/usr/bin/hadoop jar /home/user/avro-kafka/camus-wmf-0.1.0-wmf6.jar com.linkedin.camus.etl.kafka.CamusJob -libjars ${LIBJARS}"some_avro_test"  -P /home/user/avro-kafka/ >>  ./log_camus_avro_test.txt 2>&1

Revision as of 22:11, 3 August 2021