FogReplicator and Storage Nodes.
-
Server
- FOG Version: 1.3.0-RC-36 Rev 6048
- OS: Ubuntu 16.04
Description
In our environment we have 2 fog Servers, one In Northern California(SRO) and one in Southern California(MHB). The fog server in SRO runs fine, speeds are great, 3 minute laptop deployments, gigabit speed, etc. Last Friday I was working with someone in MHB showing him how to use fog and the speeds were about 100mbps. He was plugged into a switch and guessed that the switch wasn’t Gig. Later I found out from out network admin that something was saturating the link between MHB and SRO(100mb link). We found that it was the fog server. I do have a Storage Node setup from MHB that(from my understanding) is supposed to sync images from SRO. Does the storage node just tell the machine to pull images from another server in the same node? Why when I till it to pull an image that is on the local machine is it pulling from the remote machine? Should I just disable the nodes and manually sync the images from SRO to MHB? Last Friday I disabled the Storage node and today I had the network guy message me saying that it’s happening again.
Was I wrong in understanding that FogReplicator was supposed to sync images from SRO to MHB?
How can I reliably sync the images from one machine to another?
Why when I point a host at a local image is it pulling an image with the same name from the remote host?
-
How is the server setup at MHB? Is that a full fog server or a fog storage node? I’m pretty sure I already know the answer, I just want to make sure before making a left turn here.
-
They were both setup to be Full installs, I am unsure how to make one only a Storage Node.
The MHB server has a “SRO” storage node listed. Which is what I was told was required to get replication to work. Even after the “SRO” storage node had “Is Enabled unchecked” it was still pulling data from SRO. -
@sbenson OK what you are currently running is an unsupported mode of FOG. Yes it works well and how I have my fog servers setup.
In a traditional FOG setup you have one master node and one or more storage nodes. A fog storage node is exactly like a FOG master node, except a storage node doesn’t have a local database, it uses the Master Nodes database.
In your setup you have 2 fog master nodes (each with their own sql database) you have one of the servers configured as a master node and in the master node the other configured as a storage node. This IS required to get your images to flow to your remote fog server.
Now we need to understand why a client at your remote site is trying to pull an image from the HQ site. At your remote site, what values do you have for dhcp options 66 and 67. Those should point to the fog server at the remote site. In a split master setup (like you have) each site should only know about its local fog server.
-
@george1421
We confirmed that the MHB network ranges both hit the MHB server x.x.57.42. SRO is currently hosted on x.x.76.44, and the Scope in SRO reflects the same.Ideally we want SRO and MHB to be two separate servers, but sync the images between eachother. When I log into the MHB server it doesn’t show any of the images listed on SRO, it only shows the ones locally on MHB. To add to the confusion, I have added a list of ALL of our currently deployed machines to the SRO hosts. These also do not show up when logging into the MHB server.
-
@sbenson how did you create the remote (or HQ fog server) Did you clone one to the other? There is a file on the fog server /tftpboot/default.ipxe that contains the IP address of the fog server. I’m looking for cross pollination here.
In theory if you dropped the site to site link both fog servers should be able to image perfectly.
-
@george1421 Two totally separate installations. No database restorations or anything. Nothing cloned at all. Both default.ipxe’s show the correct IP addresses
-
@sbenson said in FogReplicator and Storage Nodes.:
Ideally we want SRO and MHB to be two separate servers, but sync the images between eachother. When I log into the MHB server it doesn’t show any of the images listed on SRO, it only shows the ones locally on MHB. To add to the confusion, I have added a list of ALL of our currently deployed machines to the SRO hosts. These also do not show up when logging into the MHB server.
This is fine. Its still not clear then why its using across the site to site link. If you power off the server at HQ can you image at the remote site?
-
@sbenson In a traditional FOG Master node and storage node setup you would use the location plugin to direct the clients to the right fog server for imaging. But this isn’t the case here you have to independent FOG servers.
I wonder if the spike in network traffic your network admin saw was actually image replication between the two fog servers??
-
@george1421 I haven’t tried powering off the main server. I can give that a shot and see if the MHB server even responds. I do feel this has to do with some snafu in my storage nodes. When I originally tried setting the syncing up I was under the impression it was the opposite direction(pull vs push).
-
@george1421 said in FogReplicator and Storage Nodes.:
I wonder if the spike in network traffic your network admin saw was actually image replication between the two fog servers??
This is not the case, I actually have the replication service set on a 24 hour cycle, and restart the service at 2am(iirc). This was the only way to make the replication happen at a given scheduled time.
-
@sbenson said in FogReplicator and Storage Nodes.:
I was under the impression it was the opposite direction(pull vs push).
The replication bits are isolated from the imaging components. So I don’t think a misconfiguration on this would cause imaging cross the wan link.
-
@george1421 here is a quick question. Am I supposed to have a default storage node and an extra one for SRO? This might be the problem. The original default storage node was removed so there is only a SRO storage node listed. tho the SRO server has 2 storage nodes. One labeled SyncMHB and one labeled SyncSRO. I will admit the Storage nodes have been extremely confusing from other replication services I have used in the past.
-
@sbenson ok that was confusing.
Lets say you have two fog servers (one at hq and one at remote site) and at your remote site you have a storage node (i.e. at your remote site you have 2 fog servers one is a full fog server and one is a storage node). You want images created to HQ to replicate to the other fog servers (full servers or storage nodes).
At your HQ you will create one storage group. That storage group will have your HQ FOG server configured as the default server and your fog server at your remote site. With that configuration your images from the HQ FOG server will replicate to the FOG server at the remote site. You will need to manually create the image definitions to see these images at the remote fog server.
Now at the remote fog server you create a new storage group and add your remote fog server and storage node in that storage group. The remote fog server shall not know about the HQ fog server. Its not part of its configuration. As far as the remote fog server knows, images just magically show up in its /images directory.
-
@george1421 The graphic here explains visually what I just stumbled through.
https://forums.fogproject.org/topic/6014/create-the-concept-of-a-foreignmasterstorage-deployment-node/13 -
@george1421 2 Physical servers SRO-FOG-01 and MHB-FOG-01.
for all intents and purposes SRO-FOG-01 is the Main server that will have the majority of work done on it(Creating new images, deployments, changes, etc). The MHB-FOG-01 will deploy images that I would have created on SRO-FOG-01, and synced over to MHB-FOG-01. The MHB-FOG-01 should NEVER tell anyone in the Southern California(MHB-FOG-01) network to pull from the Northern California(SRO-FOG-01) server.From what you explained it kinda matches what I have…except some things are weird.
SRO-FOG-01 has 2 storage nodes listed, SyncMHB(Not master) and SyncSRO(Master).
…SyncMHB is set to IP x.x.57.42(the IP of MHB-FOG-01), Storage Group default, Img Path = /images, etc…
…SyncSRO is set to the local IP x.x.76.44, Storage Group default, same Image path, etc.MHB-FOG-01 has one storage node, SRO(Not Master, I was told the name didn’t matter)
…SRO is set to x.x.76.44, default storage group, paths, etc…It may be best to just start over and say this. I have fog servers in 2 different cities, on 2 different internet circuits. I want the Images on Server 1 to be replicated on Server 2. What is the best method to complete this?
-
-
@sbenson said in FogReplicator and Storage Nodes.:
The MHB-FOG-01 should NEVER tell anyone in the Southern California(MHB-FOG-01) network to pull from the Northern California(SRO-FOG-01) server.
So in this case you should be able to power off the fog server at SRO and the MHB fog server and clients should image without issue. This IS the design.
-
@george1421 said in FogReplicator and Storage Nodes.:
So in this case you should be able to power off the fog server at SRO and the MHB fog server and clients should image without issue. This IS the design.
This is what we were looking for, that is why we wanted it to sync automatically during the middle of the night. Then possibly the only thing we would have to do after a sync is export the images(txt file) on SRO, and import them on the MHB server
-
@sbenson said in FogReplicator and Storage Nodes.:
SRO-FOG-01 has 2 storage nodes listed, SyncMHB(Not master) and SyncSRO(Master).
…SyncMHB is set to IP x.x.57.42(the IP of MHB-FOG-01), Storage Group default, Img Path = /images, etc…
…SyncSRO is set to the local IP x.x.76.44, Storage Group default, same Image path, etc.In this case SRO server should ONLY have the MHB fog server, with the SRO fog server setup as master. All others not the master. This will tell the Master FOG server to replicate to everyone else that is not a master.
On the MHB fog server I think you have one storage node, If that is the case create the MHB fog server as the master node and your MHB storage node as not the master.
This setup will then only sync SRO FOG server -> MHB fog server and MHB Fog server -> MHB Storage node. That way there is only one replication across the WAN.