Rolling FOG out to US Site
-
@RobTitian16 I only mentioned the script and its intended function because it would do what you needed to re-ip a host. But I agree with Wayne maybe a simpler one time run script would be in order, because sometimes you DO have to renumber a FOG server after its been setup.
-
@Wayne-Workman Many thanks, gents. It all seems to be working here in the UK so I’ll ship it out tomorrow and hopefully set up the new VM later on this week/next week.
If I run into any issues I’ll post back
-
This post is deleted! -
@Wayne-Workman Thanks for this, Wayne. Is it normal to see:
[11-01-16 9:14:18 am] | Image name: Win7Clientx86
[11-01-16 9:14:18 am] * Found Image to transfer to 1 node(s)On both the master node in the UK and the US storage node? I can’t see any received data on the US FOG server, so I’m concerned that the replication is not working correctly.
-
@RobTitian16 Is the US server a full FOG server or a storage node?
If both are full fog servers then you only configure the master server for the UK, the US fog server shall think its standalone. So you don’t setup any storage groups or anything. That is all done on the UK server.
-
@george1421 The US server is a full FOG server, but it’s listed as a storage node on the UK server. The only storage nodes on the US server are its own.
-
@RobTitian16 Are they associated to groups? If they are, is there another group they can belong to?
Replication only happens on nodes from a master->subordinate system. This is only with per group.
If an image belongs to multiple groups, though, the “primary” group will be the master and it will only distribute to other master nodes (as the master nodes will distribute to their group’s subordinates).
-
@Tom-Elliott Yes, on the UK server the images I want to replicate belong to the primary storage group called UK, and a secondary storage group called US.
The US server is listed as a storage node on the UK server, although I’ve just noticed it isn’t listed as a master node. The UK server is in its own storage group, as is the US server, so I assume all I need to do is enable the US server as a master node and it should then work?
P.S. sorry if it seems like a dumb question… I just don’t want replication to occur the wrong way/have the images wiped. -
@RobTitian16 Replication will only go Primary Master->Master between groups.
Replication will only go Master -> Subordinate within their respective groups.
I don’t know how else to make it clear.
Every group MUST have at least one “master” node. If one is not defined but there’s only one node in the group that node will “be” the master for that group (as it’s the only one available anyway).
If one is not defined and there’s multiple nodes in the group, the “Oldest” created node of that group will be assumed as the master until otherwise stated.Replication doesn’t work “up the tree” if you will.
-
@Tom-Elliott Ahh okay, thanks. I’ve changed it to the master node now for that group, so I shall wait and see what happens in that case.
-
@RobTitian16 I feel I should add some caveats to this.
The replication process is self monitored. If a replication task was started, but has not completed, the next cycle will be aware of this and will not try to replicate the file over. If replication has completed between two points and both sides have the same files the images will not be touched.
Only items (Snapins and Images) that are defined as “to Replicate” will be replicated.
Any extra data within a storage location (snapins or images) will not be removed unless that data is defined within the main system and told to do so.
For example, let’s say you decided to create a backup of image1 and you locally backed it up on a node that is not a “master”. You locally backup image1 with the name of image1_backUp within the Images storage location. image1_backUp will be untouched.
If you made a backup in the same fashion, but on a master node, that data will not be replicated to other nodes.
Only defined items will be replicated. Data loss is limited due to this implementation. At one point, FOG did used to replicate it’s images folder implicitly. This meant anything that was not a part of the “master’s” data record on another node would be removed and only the data within the master would be available on any node. This, essentially, meant that you could not maintain backup’s of things from other nodes without having to have another location available on that system. This also meant you had no granular control over what can/cannot be replicated. This is where the “Master Node” warning came from. I have not updated it because I think it’s better to be “cautious” in the case something weird does happen.
Essentially, the methods to replicate images and snapins are now much better controlled (I think) with less potential of data loss.
-
While its probably not necessairy to post this image, this is from my initial request for a multi master storage node setup. The intent was to show the relationship between the master node and the remote fog servers.
-
@Tom-Elliott Indeed, that all does make sense. Many thanks for your help! I can see the replication is going across perfectly fine.
One thing to note: I had to use the local account and the password for the ftp access, instead of the username and password listed in /opt/fog/.fogsettings (which is what it says in step 12 of https://wiki.fogproject.org/wiki/index.php?title=Managing_FOG#Storage_Management). -
@RobTitian16 The management user and password are the local linux account on the Storage nodes. If those aren’t matching something else is wrong.
The user is typically “fog” and the password is a randomly generated one during the installation which is stored in the “password” item of the .fogsettings file.
-
Given that @RobTitian16 is new to this, I’m going to say the issue is something simple.
Rob, if you aren’t using the exact password that is inside of
/opt/fog/.fogsettingsfor all the “storage nodes” you have listed in the “real” main server, the very next time you update fog, replication will break. Why? Because the FOG installer manages the local fog account, meaning the FOG installer will make sure the password for the local fog account is exactly as written in the .fogsettings file. And because you’re using a multi-master setup, you don’t have available the built-in safe-guards that the Storage Node portion of the FOG installer has in it, which already have measures to prevent this type of breakage.You should not be using the local fog account for anything, it’s very bad practice to do so. It should be reserved exclusively and only for FOG’s use. Create some other account for yourself to use.
-
@Wayne-Workman Thanks for explaining
Everything seems to be working as expected now so hopefully I’ll be seeing the new image on the US server shortly -
@george1421 Very interesting - thanks for posting! It definitely helps to see it like that.
-
@Wayne-Workman Just one final thing on this: would the replicated images on the US server show that they’ve been updated at all?
For example, if I’ve updated an image here in the UK, then see the replication has finished comparing and matching the files, should it say anything when looking at the image in the web gui (i.e. last updated) on the US server? Or do you just purely go by the replication log? -
@RobTitian16 Because the two servers each have their own DB, one DB knows about the new upload, the other doesn’t.
-
I wonder if anyone can help me with a further question I have about the US FOG server and the FOG client.
As it stands, I need the US images to have the FOG client installed which then connect to the US FOG server. However, because replication is occurring between the UK (which is the main server) and the US, it’s replicating the images with the FOG client which are configured to connect to the UK server.
I don’t want to turn replication off, but I need the FOG client on the US images to connect to the US FOG client. Is this possible at all? i.e. perhaps through a script or is there a way that FOG can install the client on a newly imaged system that’s configured for the server it pulled the image from? How do others work around this issue?
