Multiple logstash forwarder instances


(Paul) #1

Hi,

I am running multiple LF instances on a single host - unfortunately I started them both from the same working directory so they are sharing a .logstash-forwarder file.

I presume this is incorrect and I will rectify it - but could this error give rise to duplicate events that I subsequently see in Elasticsearch ?

Thanks,


(Magnus Bäck) #2

Are both instance reading from the same file? If so then yes, it could definitely lead to duplicate events.


(Paul) #3

Hi,

Thanks for your reply.

No, both instances are reading files from separate directions but they are sharing the same .logstash-forwarder file.


(Magnus Bäck) #4

Hmm. That probably results in them periodically trampling on each other's .logstash-forwarder files, so if you restart both then one of them is going to lose its state. You should grab a copy of the file, shut down the instance that created the file, shut down the other one after you've made that it has flushed its internal state and overwritten the state file, then restart the LSF instances from different directories to which you've moved each instance's state file.


(Paul) #5

Thank you again for your reply. I have refactored my scripts to ensure that each LF instance now starts in a unique working directory.

BTW, I also found a workaround for this issue by using the logstash fingerprint filter to get a hash of the message and then using the fingerprint value as a document_id when submitting the document to elasticsearch. This ensures no exact duplicates make it into elasticsearch.


(Yarden Bar) #6

Hi Magnus,
Can you ellaborate on how to do so?
In my use-case, LSF is installed from rpm to /opt/logstash-forwarder and is started with init script (/etc/init.d/logstash-forwarder)

I don't understand how shall I start it...

Thank you,
Yarden


(Magnus Bäck) #7

Can you ellaborate on how to do so?

Do what? Did you also manage to start two instances of LSF from the same directory?


(Yarden Bar) #8

Yes, but I suspect they clash because they both use /var/lib/logstash-forwarder/.logstash-forwarder as a state file.

I understand that I should stop using LSF init script and run it from different directories? this means that the state file gets created in the directory I'm running the command from?


(Magnus Bäck) #9

Yes, LSF creates the state file in the current directory. That doesn't mean that you need to stop using the init script, but you may have to modify it.


(Yarden Bar) #10

I've modified the init script to reflect the second LSF.
Also modified /etc/default/logstash-forwarder-OTHER file, configured 'chdir' to a different directory.

Should I expect to see the second LSF state file in 'chdir' directory?

I'm using lsof to determine which state file each LSF processes is using.


(Magnus Bäck) #11

If you succeed in changing the starting directory of the LSF process then that's where you'll find the state file.


(Yarden Bar) #12

I think that I've manage to change the starting directory:

sudo pwdx 30315 # LSF-OTHER PID
30315: /var/lib/logstash-forwarder-OTHER

But I don't see the state file there:

ls -la /var/lib/logstash-forwarder-OTHER
total 8
drwxrwxr-x  2 logstash-forwarder logstash-forwarder 4096 Nov  4 08:03 .
drwxr-xr-x 27 root               root               4096 Nov  4 08:03 ..

The server is "Scientific Linux release 6.1 (Carbon)"
Also, nothing in the logs to indicate problem with opening/creating the state file.


(system) #13