Trying to setup a graylog cluster on AWS to use for all my logging, but I'm 
having trouble getting it to work - surely lots of people have this problem 
but I can't find any docs/forum posts about it!

The documentation has bits that talk about various things, and the 
graylog-ctl should help a lot, but it clearly misses out all sorts of 
important areas that need sorting!  Can anyone help?

What I've got is:

graylog-web - a web-interface server
graylog-a - a "server", pointed to -b as the "master"
graylog-b - a "server" as well - the master, and the original server
graylog-data-a - a "datanode"
graylog-data-b - a "datanode"

Problems are as follows:

- There's no apparent way to know what's happening with the elasticsearch 
cluster, except looking at the /system/indices bit.  This is OK so long as 
it's green, but otherwise how are you to know where a problem lies?   It 
seems for example that if I start data-a before data-b, then they don't 
connect as a cluster and it goes yellow with 6 shards active and 6 
unassigned.  Restarting data-a then fixes the problem, but it seems that 
one of them is a master and needs to start first or something?

- It was all working OK with just one server node, but then I added a 
second node (in this case, graylog-a).  This screwed everything up, 
basically.  I simply created a new instance, ran the sudo graylog-ctl 
set-cluster-master x.y.z command, then reconfigure-as-server, and all 
should have been well you'd think.  

I realised that although it configures to talk to mongodb on the master 
server, mongo isn't actually clustered between the servers, so I think 
there's still a single point of failure here?  I added the relevant port to 
the firewall and initially all was well, but I assume that if graylog-b 
(the master) goes down, then the whole thing will.  Shouldn't the 
graylog-ctl command handle setting up mongo as a cluster as well, or is it 
not that easy?

- I then wanted to change the way the indices are setup, so I ran the sudo 
graylog-ctl set-retention --size=5 --indices=15 command on the master node, 
but it's not really clear whether this needs to be run there, or on the 
elasticsearch node(s), or everything?  I think the config is stored in 
mongo, but then it's also in the .conf files??  So I ran it on the second 
server node, and the whole thing exploded.  I got some weird error trying 
to log in, which I think was caused by the graylog.conf file having 
different root_user names set between the two servers (this is clearly not 
copied when you do the cluster-master bit at the start). Having fixed that 
it then still wouldn't load, but a restart of everything seems to have 
brought it back together.



Argh!  All quite complex and would be helpful to have a guide that just 
explained how it might work and what commands to run on what machine/what 
you need to do manually.

In short, its a nice idea to have the clustering setup using the tools, but 
it's so hard to know what's going on and there's so little actual 
documentation, it's a bit useless at the moment - yet could be so good no 
doubt quite easily!

thanks
Pete

-- 
You received this message because you are subscribed to the Google Groups 
"graylog2" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to