Primary + SDS cluster failvoer through cm
Is there a way to tell the connection manager to try to start a node that is down? I dont mean the "turn the sds to new primary y old primary is down" that already does it, I mean a "try to start the primary that failed as an sds node of the new primary you just started". I don't see anything like that anywhere in the documentation. I guess it must be done through alarmprogram.sh o cmalarmprogram.sh... If this is the case, someone tried this? I'm not sure if the cm should be the one that try to wake up the fallen node or perhaps the just made primary node must try to wake up the sds nodes...
Are you sure that this would be a good idea? Generally when a server goes down there is either a person that brought it down or there is something that caused it to go down. With SDS that would generally be an environmental reason such as a machine failure, power failure, or network failure. If a person manually brought it down, it would be a nuisance for the system to automatically bring it back up. And if the system went down, wouldn't it make sense to first fix the problem in the environment that cause it to fail before bringing it back online? Madison Pruet Retired and Loving it On Tuesday, May 3, 2016 9:17 AM, JACOBO BALBUENA <jacobo.bc@gmail.com> wrote: Is there a way to tell the connection manager to try to start a node that is down? I dont mean the "turn the sds to new primary y old primary is down" that already does it, I mean a "try to start the primary that failed as an sds node of the new primary you just started". I don't see anything like that anywhere in the documentation. I guess it must be done through alarmprogram.sh o cmalarmprogram.sh... If this is the case, someone tried this? I'm not sure if the cm should be the one that try to wake up the fallen node or perhaps the just made primary node must try to wake up the sds nodes... ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.
Guess they are asking me this to prevent night phone calls... Anyway I'm only willing to force the wake up of sds and connetion managers that are down, don't want to get close to split brain problems, and always leaving a room to "not wake up things" through script...
You can avoid the split brain problem on SDS by using alternate communication through the blobspace blob communication feature. Madison Pruet Retired and Loving it On Wednesday, May 4, 2016 7:37 AM, JACOBO BALBUENA <jacobo.bc@gmail.com> wrote: Guess they are asking me this to prevent night phone calls... Anyway I'm only willing to force the wake up of sds and connetion managers that are down, don't want to get close to split brain problems, and always leaving a room to "not wake up things" through script... ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.
What might happen, however, is that you could end up with a never-ending cycle of up-down-up-down.... Madison Pruet Retired and Loving it On Wednesday, May 4, 2016 9:13 AM, Madison Pruet <madison_pruet@yahoo.com> wrote: You can avoid the split brain problem on SDS by using alternate communication through the blobspace blob communication feature. Madison Pruet Retired and Loving it On Wednesday, May 4, 2016 7:37 AM, JACOBO BALBUENA <jacobo.bc@gmail.com> wrote: Guess they are asking me this to prevent night phone calls... Anyway I'm only willing to force the wake up of sds and connetion managers that are down, don't want to get close to split brain problems, and always leaving a room to "not wake up things" through script... ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.