Skip to main content

Centralized logfile management across networks with flume


Facebooks's scribe was the first available service for managing a hughe amount on logfiles. We didn't talk over 2 GB / day or so, I mean more as 1 TB per day. Compressed.
Now, a new apache incubator project is flume [1]. It is a pretty nice piece of software, so I love it. It is reliable, fast, safe and has no proprietary stack inside. And you can create really cool logging tasks.

If you use Clouderas Distribution you get flume easy with a "yum install flume-master" on the master and "yum install flume-node" on a node. Check [2] for more infos about.

Flume has a lot of sources to get logfiles:
- from a text-file
- as a tail (one or more files)
- syslog UDP or TCP
- synthetic sources

Flume's design belongs to a large logfile distribution process. Let's assume, we have a 100 Node Webcluster and incoming traffic around 3 GB/s. The farm produce 700 MB raw weblogs per minute.
Through the processing over flume we can compress the files, sort them into buckets you need and fast deliver into our hdfs. Here a working example:

cat /flume-zk/running.cfg
collector1.local : autoCollectorSource | collectorSink( "hdfs://namenode.local:9000/user/flume/weblogs/%Y-%m-%d/%H00/%M/", "%{host}-" );
collector2.local : autoCollectorSource | collectorSink( "hdfs://namenode.local:9000/user/flume/weblogs/%Y-%m-%d/%H00/%M/", "%{host}-" );
collector3.local : autoCollectorSource | collectorSink( "hdfs://namenode.local:9000/user/flume/weblogs/%Y-%m-%d/%H00/%M/", "%{host}-" );
collector4.local : autoCollectorSource | collectorSink( "hdfs://namenode.local:9000/user/flume/weblogs/%Y-%m-%d/%H00/%M/", "%{host}-" );
agent1.local : syslogTcp( "19800" ) | autoE2EChain;
agent2.local : syslogTcp( "19800" ) | autoE2EChain;

The chain autoE2EChain describe a failover process, if one of the nodes didn't respond they will be moved at the end. You will see the logical mapping at the webinterface (http://flume-master:35871/flumemaster.jsp).
We split the data here into minutes and set as a identifier the host from which we get the logs at the end. Make it easier to debug. The webfarm logs via a loadbalancer to the agents (input syslog, output 19800). 19800 is a free unprivileged port, as example.

Let us check one of the agents:
# cd /tmp/flume/agent/agent1.local/
# ls
done  logged  sending  sent  writing

# ls -lah writing/
total 418M
drwxr-xr-x 2 flume flume 4.0K Sep 28 16:25 .
drwxr-xr-x 7 flume flume 4.0K Sep 28 10:21 ..
-rw-r--r-- 1 flume flume 418M Sep 28 16:25 log.00000019.20111011-162503316+0200.10204828793997118.seq

That the logfile we receive at the moment from our nodes. After writing (you can define the split in flume.conf) the log will be sent to the collectors, so we connect to collector1:

# tail -f -n 10 /var/log/flume/*.log
<del>: Creating compressed HDFS file: hdfs://namenode.local:9000/user/flume/weblogs/2011-09-28/1600/25/agent1.local-log.00000019.20111011-162503316+0200.10204828793997118.seq.bzip2
<del>: Finishing checksum group called 'log.00000019.20111011-162503316+0200.10204828793997118.seq'
<del>: Checksum succeeded 1325c55440e
<del>: moved from partial to complete log.00000019.20111011-162503316+0200.10204828793997118.seq
<del>: Closing hdfs://namenode.local:9000/user/flume/weblogs/2011-09-28/1600/25/agent1.local-log.00000019.20111011-162503316+0200.10204828793997118.seq
<del>: Closing HDFS file: hdfs://namenode.local:9000/user/flume/weblogs/2011-09-28/1600/25/agent1.local-log.log.00000019.20111011-162503316+0200.10204828793997118.seq.bzip2

You see, the collector receive the action from the agent, open a sink into hdfs, write via stream the file and close the sink after the time we configured. Pretty nice! The logging mechanism works perfectly, the files will be splitted and compressed as bzip into 1 minute-pieces into our hdfs. Remember, use always bzip as compress codecs because the codec understand markers (blocksize, reducing etc).



Popular posts from this blog

Deal with corrupted messages in Apache Kafka

Under some strange circumstances it can happen that a message in a Kafka topic is corrupted. This happens often by using 3rd party frameworks together with Kafka. Additionally, Kafka < 0.9 has no lock at at the consumer read level, but has a lock on Log.write(). This can cause a rare race condition, as described in KAKFA-2477 [1]. Probably a log entry looks like: ERROR Error processing message, stopping consumer: ($) kafka.message.InvalidMessageException: Message is corrupt (stored crc = xxxxxxxxxx, computed crc = yyyyyyyyyy Kafka-Tools Kafka stores the offset of every consumer in Zookeeper. To read out the offsets, Kafka provides handy tools [2]. But also can be used, at least to display the consumer and the stored offsets. First we need to find the consumer for a topic (> Kafka 0.9): bin/ --zookeeper management01:2181 --describe --group test Prior to Kafka 0.9 the only possibility to get this inform

Hive query shows ERROR "too many counters"

A hive job face the odd " Too many counters:"  like Ended Job = job_xxxxxx with exception 'org.apache.hadoop.mapreduce.counters.LimitExceededException(Too many counters: 201 max=200)' FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.MapRedTask Intercepting System.exit(1) These happens when operators are used in queries ( Hive Operators ). Hive creates 4 counters per operator, max upto 1000, plus a few additional counters like file read/write, partitions and tables. Hence the number of counter required is going to be dependent upon the query.  To avoid such exception, configure " mapreduce.job.counters.max " in mapreduce-site.xml to a value above 1000. Hive will fail when he is hitting the 1k counts, but other MR jobs not. A number around 1120 should be a good choice. Using " EXPLAIN EXTENDED " and " grep -ri operators | wc -l " print out the used numbers of operators. Use this value to tweak the MR s

GPT & GenAI for Startup Storytelling

OpenAI and Bard   are the most used GenAI tools today; the first one has a massive Microsoft investment, and the other one is an experiment from Google. But did you know that you can also use them to optimize and hack your startup?  For startups, creating pitch scripts, sales emails, and elevator pitches with generative AI (GenAI) can help you not only save time but also validate your marketing and wording. Curious? Here are a few prompt hacks for startups to create,improve, and validate buyer personas, your startup's mission/vision statements, and unique selling proposition (USP) definitions. First Step: Introduce yourself and your startup Introduce yourself, your startup, your website, your idea, your position, and in a few words what you are doing to the chatbot: Prompt : I'm NAME and our startup NAME, with website URL, is doing WHATEVER. With PRODUCT NAME, we aim to change or disrupt INDUSTRY. Bard is able to pull information from your website. I'm not sure if ChatGPT