-
Notifications
You must be signed in to change notification settings - Fork 5.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Multi-master syndic is not documented in all examples hot-hot, multimaster and multi syndic w multimaster #18751
Comments
That's fair... ;) I can probably get that put together over the next week
|
@jacksontj I'm happy to get that added too...but I don't know how. On Thu Dec 04 2014 at 5:44:58 PM Thomas Jackson notifications@github.com
|
I'm quite interested in this, as well! |
I misunderstood this addition. I thought it was to have a single syndic connected to two hot master-masters (e.g., how vanilla multi-master works). @cro set me straight. Would still be nice to have docs, but I see now there are no new config options for this. |
@whiteinge Yea, this is basically just a way for you to list out a bunch of "syndic_master"s. There are a few caveats in the current implementation (which is what I'd like to document), but as you said there are no new config options. |
Thank you for the clarification! On Fri, Dec 5, 2014, 9:19 AM Thomas Jackson notifications@github.com
|
@whiteinge would you mind clarifying your clarification? |
@UtahDave A syndic cannot currently have a connection to multiple masters. This ticket is for the following scenario which provides more redundancy than just a master -> syndic-master connection but does not provide full HA. Given Master A and Master B at main HQ, and syndic-masters a & b in one data center and syndic masters c & d in another data center:
If either Master A or Master B goes down, you still have a connection left to at least one syndic in each data center. But if the wrong syndic goes down at the same time as the wrong master, then you can still lose connection to a data center. |
To clarify more, since the addition of MultiSyndic you can put a list (instead of a single string) as the master list for a syndic-- and the syndic will connect to all of them-- similar to multiminion. The code is a bit messy, but it does work ;) |
We need to completely overhaul the topology section of the docs to explain Salt's deployment options: http://docs.saltstack.com/en/latest/topics/development/topology.html |
+1 |
Holy cow, @jacksontj, that's awesome! I was previously told multi-connections had specifically not been added. Definite +1 to getting MultiSyndic documented and also adding this to the topology doc. |
The only reason I haven't pushed it a lot is that some of the failure cases are a bit odd... I might just go back and thread it to work around them. Basically in a failure scenario (one master unavailable) you have to let zmq attempt to re-connect and we have to re-auth through it-- which means that the syndic will lock up for however long that connect timeout is. In general slow is better than unavailable-- but it can be worked around, I'll try to get back to that after our 2014.7 release internally |
Odd failure cases with ZeroMQ, you say? Hm. ;-) Thanks for mentioning that. I can think of one or two large sites just offhand that will be interested in getting this set up. |
We are one of them ;) On Wed, Feb 18, 2015 at 8:33 AM, Seth House notifications@github.com
|
Hi, @jacksontj, I setup 4 syndic-master that the one would connect to the other three, just like #6 arch you introduced:
After all setup, I fired one job on syndic01, and I actually find out a duplicate of event loop problem, syndic 02\03\04 will duplicate to forward job and re-publish to its localmaster. so it will be:
I checked out the MultiSyndic code, it seems that you only not forward event when the After I check the 2014.7.5 & 2015.5.0 version, the setup cluster still doesn't work, still loop、loop and loop out... Help~~~~~ |
so multisyndic is not working as it suppose to? |
Yeah, Multi-Master with Multi-Syndic ( cluster mode ) didn't work in my test case, but Multi-Syndic does. |
@jacksontj Multi-syndic setup has missing return issues from master of master for me. i have master_id set at the both master nodes and have a list of masters listed as syndic_master on the syndic nodes. Am i missing anything?? Seems like increasing the number of timeout fixes this problem - I did noticed that all the ret events arrive accordingly in each master of masters, but not all returns are showing up in --out-file but if i look up the JID afterwards, it has correct number of returns. |
I'm having a super interesting issue with this set up actually. I have the following servers:
When I run a test.ping on master1 I get the following:
Yet when I run the same on master2, I get the following:
You'll notice on master2, syndic2 isn't responding, despite the key being accepted and transferred correctly. The syndic logs keep having the same message repeating with regards to master1:
When I start the salt-syndics from scratch, keys are exchanged on both master1 and master2 however?
Digging down, I found that when I started the minions, syndic1 has the keys sent to it, but syndic2 does not, despite the configuration being the same on both minions with regards to having the syndic pub keys on them, here's what my syndic1
syndic2
Here's what i have for my config files on the syndics for the masters:
And here's what I have on the minions for the syndics:
Going a bit deeper, I can definitely confirm on the network level all instances should be able to talk each other: Here's an example telnet from syndic2->master1
And for those counting at home master1->syndic2
I know it's a lot of information so I'm happy to confirm anything needed from above, but I REALLY need a setup like this to work with syndics and masters for HA if I'm going to further proceed with SaltStack. |
Make sure to have master_id set on the mom if you have not yet. With
|
@mimianddaniel thanks for the suggestion, just gave that a go and still in the same spot. That was something I meant to have set anyways so at least I've got it covered now ;) |
Been over a month since any activity on this thread, just wondering if anyone from SaltStack has been able to plug at this. |
Doesn't look like we've had time to get to documenting this yet. @jacobhammons Can you have a look here when you get a moment? |
Had unexpected issues pop up, this one has to slip to the next release |
Closing as Won't fix. Please follow Syndic deprecation work https://github.com/orgs/saltstack/projects/42 due for 3007.0 |
@anilsil I get 404 for https://github.com/orgs/saltstack/projects/42 |
Still in WIP but you can refer this #64941 |
As the title says. I can't help but feel this is somehow @jacksontj's fault! ;-)
The text was updated successfully, but these errors were encountered: