Saddly, during the reimage of all elasticsearch servers, we did not use a no swappartman configuration. We should disable swap nonetheless (to make elasticsearch happy about memory locking configuration) and ensure that the next reimage will not use swap.
The following should be sufficient:
swapoff -a sed -i.bak '/swap/d' fstab
This does not recover the 1Go of the swap partition (but we don't really care about 1Go). We do not manage swap via puppet, so it does not seem to make sense to apply that change via puppet.
The partitioning should be updated to use raid1-lvm-ext4-srv-noswap.cfg for the next reinstall.
@RobH do you have an opinion on the above plan?
This all makes sense to me. I agree that implementing the swap removal via puppet is likely overkill, since this is a clean up and not expected to be an ongoing item to be managed by puppet. Removing the swap but not re-gaining the 1GB seems better than just leaving it in place, and simplifies fleet wide partitioning.