HA Deployment: second controller

[lgretton]

The documentation is a bit sparse about how to install TX for a secondary controller. In our case the prepare script fails as 'hostname -s' does not match the value of trix_ctrl1_hostname in all.yml, but that's because we are trying to install what should be matching the trix_ctrl2_hostname value, but there's no check being done for that within prepare.sh.

So at the moment prepare.sh fails with the following error which we can't get past:

****************************************************************************
*                                                                          *
*  Please note the hostnames are not matching (see                         *
*  site/group_vars/all.yml).                                               *
*  Please configure the network before starting Ansible                    *
*                                                                          *
****************************************************************************

hostname -s returns 3x-head02. In our all.yml file we have these options, having simply copied the all.yml file over from the first controller.

trix_ctrl1_hostname: 3x-head01
...
trix_ctrl2_hostname: 3x-head02

It feels like we're missing an argument to the installation TUI script to specify that the installation is for a secondary controller.

Also we're not sure whether the shared /trinity disc should be mounted first on the secondary controller, or should the installation of the secondary controller take care of that for us?

[aphmschonewille]

Hi Igretton

There is a bug in prepare.sh to verify a matching hostname. It's harmless though but will be fixed in the next release.

Trinityx takes responsibility for the shared disk, including the mounts and creating file systems when using drbd. It would suffice to assign a empty suitable disk and set the preferred layout. Zfs could be your friend. More info about the definition of the shared disk amd how trinityx utilizes it can be found here https://docs.clustervision.com/install/shared_fs_disks/

One note. Recently we found that not all kernels are suitable to use drbd. This means that in some cases a specific kernel should be used when using drbd (kernel modules). We didn't publish a matrix as this is very fluid.

Setting up HA can be challenging i admit, but if you have the right kernel and configure the shared disk definition in all.yml appropriately, the install goes without issues.

The installation for the second ctrl is easier than the first where the second wont setup the shared disk. It merely configures it, assuming the first ctrl config is done and all has been taken care off. I prefer to install ctrl1 first, then copy the all.yml to the second ctrl, tune maybe the disk/device for the shared disk (i.e. uuid disks are unique) and let it run as well. No ssh keys sharing needed.

The decission for who is first or second is based on ctrl1 vs ctrl2 configs in the all.yml file. I.e. if a hostname -s matches trix_ctrl2_hostname, it will be seen as second. There is a mechanism included that uses pcs info as well, in case you have the second being active and you want to reinstall the first. This however is not in order for your case.

Feel free to reach for more info if needed.

-Antoine

[lgretton]

Thank you Antoine. I think we had probably not configured the shared_fs_disks properly. We're moving on to another attempt with a disk device on shared storage with ZFS anyway, so this issue can be closed.