May I use Collector Group for Server Fail Over?
Hi There,
I see the new function of dT6 regard collector group that allow me to setup fail over among difference collectors. My question is if I can use this function for dT server fail over? The scenario is as above,
-Agents connect to single configured collector in collector group.
-Collectors in one collector group are report to two difference server
-One dT Server goes down, can agent continue work with other collectors which report to another dT server in sames collector group?
Thanks
YC
YC,
no you can't do that. Collectors cannot report to more than one server.
Reinhard
Hi Reinhard,
I meant I include difference collectors in one collector group. For example , I have collector A and collector B belongs same collector group. Collector A report to Server A, Collector B report to Server B. If Server A goes down, can the agents report to collector A switch to collector B. I just looking for a HA solution of dT server within dynaTrace technology.
Thanks
YC
Still you can't do that. Collectors that might report to different servers might also have different configurations from those servers.
The collector will now know that it's server is down and tel the agent to connect to another collector that is configured on a second server.
I know we had a few customers that did failover on a lower level for the dynatrace server (IP failover) but the dynatrace server per se doesn't have HA capabilities (considering session filestorage that would need to be accessed r/w by two servers, database access etc..)
Reinhard
Thanks at first.
If the servers are configed identically and we are OK with losing some historical data in session, then the server fail over is not possible?
Best
YC
Hi YC,
You can configure dT server failover, but as Reinhard mentions this happens at a low level, and has nothing to do with dynaTrace. It is a somewhat tedious task that has to be done by someone who knows how to set up active/passive clusters in something like MSCS. At a high level, it would go like this... you'd have two servers, A and B. Only one would be active at a time. There would be a VIP established pointing to the active server. The easiest (simplest) way to go about it would be to have the dynaTrace home directory on a SAN, able to be shared by both servers A and B. That would ensure that all critical config files, etc, were shared by each server. Also, the session storage would need to be on a SAN as well, again sharable by both servers.
The trick would be identifying to the cluster software what it means for the dT server to be down on the primary (active) server, so that it could fire up the dT server process on the passive server and re-point the VIP to the newly active server. As Reinhard mentioned, dT has no inherent HA functionality or cluster-awareness.
This should work for you if you pay attention to all the clustering details. dynaTrace would have no idea that you are doing this.
Rob
I have collectors group setup and when the main collector that has been used in Monitors settings goes down all monitors stop working till I configure them on-line collector instead.
Please advise.
I don't think it is possible to have fail over for a monitor collector. Agents can fail over between collectors in a group, but monitors have to be set up to run on a single collector at a time and there is no "standby" arrangement possible at this point. .
Thank you, Dave. It is not very good news. If collector goes offline all our monitors won't work.
A best pratice is to install separate collectors that execute monitors. Separate from those collectors that capture data from agents. With that you ensure that your collectors are not impacted by a situation where a lot of new agents are starting and consuming a lot of CPU and with that maybe impacting monitor execution.
Also - if a collector looses the connection to the server it will buffer the data. Once the connection is re-established the data will be sent. In that respect you dont often really loose data
You might be able to create a "hot standby" collector that could be activated via a script that pushes the appropriate configuration files from some shared repository and starts the "standby collector". Just brainstorming here...
Community Member of the Month
April 2018
Announcing Dynatrace's Community Member of the Month for April 2018, Babar Q.! Click here to read more! 