Forum
Welcome, Guest
Username: Password: Remember me

TOPIC:

XenServer 6.5 6 years 8 months ago #354

  • Salvatore Costantino
  • Salvatore Costantino's Avatar
  • Offline
  • Posts: 703
Hello Bob,
Our storage framework (iscsi-ha) includes pre-compiled DRBD binaries for all recent kernel versions of XenServer. You can find the binaries in the /RPM/ folder of our tarball ** However ** if you are building your cluster on top of XenServer 6.5, then a kernal module is not required as DRBD is part of the newer kernel used in XenServer 6.5. You will still need to build and compile the userland tools for DRBD - But, that too is not necessary since we include pre-built binaries in our distribution tarball.

If you follow our How-To document for building the cluster, be aware that there are some slight changes required when building on top of XenServer 6.5. The Docs should be updated in the coming weeks. In the meantime, you can try a new installer which will build the entire cluster for you.

All you need to do is:
- build a XenServer pool with 2 hosts (1 master and 1 slave)
- create dedicated replication network interfaces between the servers and connect directly via an ethernet patch cord.
- assign an IP address to the replication interfaces. If you want to keep it simple, use the default values from our installer.. 10.10.10.1/24 will be assigned to the master's replication port and 10.10.10.2/24 will be assigned to the slave's replication port. The assignment of IPs can be done from within XenCenter
- Download the installer and run it on both hosts..

That's it.
The intaller will install the proper versions of DRBD modules based on your base XenServer version.

The installer can be downloaded here:
halizard.org/release/noSAN-combined/hali...osan_installer_1.4.2

Please Log in or Create an account to join the conversation.

XenServer 6.5 6 years 8 months ago #355

So, I ran the installation and it installed great on my slave. I am currently waiting for my backups to be done so I can install it on the Master. Now, I was unable to connect to the iSCSI storage repository, but I am assuming that is because I don't have the Master setup yet. Correct me if I'm wrong. :)

Now the installer asked for a heuristic IP address. What exactly does that mean. In my setup, I made the "Internet/Management" interface have a different IP Network than the Replication interface. Is that going to matter?

Lets say (These are fake)
Management/Internet = 192.168.1.*
Replication = 10.10.10.*

Is that going to matter? Especially if I point the SR to 10.10.10.3

Thanks for your help on this, you rock!

Please Log in or Create an account to join the conversation.

Last edit: by Bob.

XenServer 6.5 6 years 8 months ago #356

  • Salvatore Costantino
  • Salvatore Costantino's Avatar
  • Offline
  • Posts: 703
you are correct, the iscsi-sr IP of 10.10.10.3 will appear automatically on your master once the installation has been completed..

making the management IP differnet than the replication is actually preferred - however, it will work either way. Ideally the replication network should be completely closed and directly attached via cable. So, having management on 192.x.x.x and replication on 10.x.x.x is perfectly fine

regarding the heauristic IP - this is an important setting required for HW high availability (will not work without it). you can read up on it in our ha-lizard documentation. link to docs below:
halizard.com/support-for-ha-lizard/documents-for-ha-lizard

Please Log in or Create an account to join the conversation.

XenServer 6.5 6 years 6 months ago #388

Hello,

i just tried to install HA-Lizard using the new script and it bails out at the end of installation. Please look below for log.

The setup consists of 2 x HP Proliant 380 with 1 Raid-1 local storage each. The bonding interfaces are up and running.

The local storage on my pool master server is "toast" btw. It had been removed by the script.

Question: Must i somehow reverse the changes the script did make (i assume theres no rollback) before i can try the script again?

Thx & Cheers,

Robin
Installed:
  scsi-target-utils.x86_64 0:1.0.14-2.el5                                                     

Dependency Installed:
  libibverbs.x86_64 0:1.1.6-3.el5           librdmacm.x86_64 0:1.0.15-2.el5                  
  openib.noarch 0:1.5.4.1-4.el5             perl-Config-General.noarch 0:2.40-1.el5          

Complete!
Convert local storage to iSCSI-HA storage? <yes or no>yes
Hostname detected = POT-XEN-1
Local Storage SR UUID discovered = e86b1a1a-9e97-65dd-f1b6-75535a036489
Local Storage PBD UUID discovered = 36ae0f17-c0a4-a7d8-d71b-e092c3df246e

Press Enter to Continue


Unplugging PBD 36ae0f17-c0a4-a7d8-d71b-e092c3df246e

Destroying PBD 36ae0f17-c0a4-a7d8-d71b-e092c3df246e

Removing Storage e86b1a1a-9e97-65dd-f1b6-75535a036489
Local storage repository has been successfully removed
Backing block device set to /dev/sda3

Enter LUN to be used for exposing iSCSI storage and press <Enter>
(Leave blank to accept default value of 10)
LUN set to 10

Backing up LVM configuration file /etc/lvm/lvm.conf to /etc/lvm/lvm.conf.halizard_restore
  Reading all physical volumes.  This may take a while...
  No volume groups found
./halizard_nosan_installer_1.3.8: line 800: /etc/iscsi-ha/iscsi-ha.conf: No such file or directory
Enter heuristic IP address used by HA-Lizard.
This can be any IP address reachable by this host
by traversing the XenServer management network and press <Enter>
Heuristic IP set to 

Press <enter> to continue

Invalid IP Address Specified !   Exiting...

Please Log in or Create an account to join the conversation.

Last edit: by Robin.

XenServer 6.5 6 years 6 months ago #389

  • Salvatore Costantino
  • Salvatore Costantino's Avatar
  • Offline
  • Posts: 703
Hi Robin,
There is a rollback feature in the installer that we put in for development purposes.. You can give it a try to bring your system to its original state. It should restore your original config files and recreate the local storage. The roll back is a bit crude and you may see some errors but it gets the job done.

The installer you ran should have put some data in /tmp/halizard_tmp_/
In this folder you should see at least one folder with a UUID as its name. Take a look inside the directory with the UUID name and ensure that that your original iptables configuration and lvm.conf are there. If OK, then run the installer again, only this time passing in the UUID (directory name) as an argument.
ex: "./halizard_nosan_installer_1.3.8 7b0718ce-d337-48c2-b5e7-313e4a6b1df4"

Once you get the system to its original state, you can proceed to run the installer again. We have a newer installer available.. You can check out this page to grab the latest one:

halizard.org/release/noSAN-combined/

Regarding the error you encountered. The installer was expecting an IP address for the heauristic address used for HA (this is mandatory in a 2 node pool as it is used to prevent split brain). Perhaps we can make the installer more fogiving by forcing an input rather than exiting the script. Regardless, when you run the installer again, make sure to enter an IP address when prompted. The heuristic IP should be an IP on the management network as close to the servers as possible.. Maybe the IP address of the ethernet switch the hosts are connected to. The configured IP should also respond to ping. That should get you past the error.

Please Log in or Create an account to join the conversation.

XenServer 6.5 6 years 6 months ago #391

Hello and thx for the help.

I did re-install the servers with the first 2 patches for 6.5 and did run the scripts again. The script didnt bail out but the iscsi interface didnt come up.

I could track the problem down to the drbd service not beeing propery configured because the host-names of the maschines had some upper-case chars which resultet in drbd not starting. (i.e. XenServer) After changing the host and service names to all lowercase i had to manually setup drbd and after this i could connect my pool to the iscsi storage.

EDIT: I had performance issues but Performance seems to have increased. I did remove the according text and logs

I just discovered this in the daemon.log
Apr  9 19:19:27 xen-2 multipathd: drbd1: add path (uevent) 
Apr  9 19:19:27 xen-2 multipathd: drbd1: failed to get path uid  

The entries are 1 (slave) and 2 (master) hours old. Is this normal?

Also interesting would be what is the best way to verify data integrity?

Again, thx a lot for your help!

Cheers,

Robin

Please Log in or Create an account to join the conversation.

Last edit: by Robin. Reason: Seems like the performance increased after applying the last 4 patches and rebooting!