hello Phil
you say that the two cluster node can't talk together ? In this case yes your setup is complex :).
For your information drbd is not shared storage. It's network block replication. For exemple we gto 12k devices and 20mbps network traffic for drbd replication. When we switch we got exactly the same information on the second node. So we can ack, maintenance etc.
this is very efficient.
oau
----- Mail original ----- De: "Phil Wild" <phil at holobyte.com.au> À: "Olivier AUDRY" <olivier at audry.fr> Envoyé: Jeudi 31 Mars 2011 10:14:59 Objet: Re: [Xymon] xymon and couchdb
Thanks,
I can't provide a heartbeat interface between the two xymon servers and there is no shared storage available, now can I fail an IP address over between the nodes. We have a 5 minute life on the DNS CNAME for the xymon server and fail over from one site to the other by switching the CNAME. Configuration files are synchronised from live to standby host using rsync. Both xymon servers pick up faults and both try to send out messages. All alerts go out through a SCRIPT (both sms and email) and the script drops the alert if the server is the standby. The servers know if they are active or standby by checking the content of a text file prior to delivering the message. If the text file contains the word active, messages are delivered, if the word standby is detected, the messages are dropped.
Acking an alert becomes a little more complex.
Cheers
Phil
On Thu, Mar 31, 2011 at 3:20 PM, Olivier AUDRY < olivier at audry.fr > wrote:
hello Phil
for a good auto failover we use heartbeat and drbd. Right now is working very well and the implementation is very simple et efficient. 5s to switch.
oau
----- Mail original ----- De: "Phil Wild" < phil at holobyte.com.au > À: xymon at xymon.com Envoyé: Jeudi 31 Mars 2011 08:54:52 Objet: [Xymon] xymon and couchdb
Hi Xymon users,
I have a question and was wondering what people thought. We currently run two xymon servers to provide HA (servers are in different data centres with only IP connectivity between them). All clients update both servers and we replicate the state file at regular intervals to help with failover (manual) in the event of an issue with the servers.
We were looking at automating the failover process and the subject of Apache CouchDB was raised. Reading through some documentation on couchdb, I thought that perhaps couchdb would make a good backend repository for xymon state information.
I was wondering if anyone has any experience with couchdb from a performance perspective?
It would be a big change to xymon to use couchdb as the repository for xymon state but it may aid in allowing for a more distributed/clustered xymon deployment.
Thoughts?
Phil
Xymon mailing list Xymon at xymon.com http://lists.xymon.com/mailman/listinfo/xymon