From d835b2cae8abc71958b69362162e6a70c3d7ef63 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Wed, 17 Apr 2024 08:48:59 +0200 Subject: Adding upstream version 4.6.0. Signed-off-by: Daniel Baumann --- doc/bootstrap-todo.md | 56 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 56 insertions(+) create mode 100644 doc/bootstrap-todo.md (limited to 'doc/bootstrap-todo.md') diff --git a/doc/bootstrap-todo.md b/doc/bootstrap-todo.md new file mode 100644 index 0000000..3db7136 --- /dev/null +++ b/doc/bootstrap-todo.md @@ -0,0 +1,56 @@ +# Bootstrap TODO + +(inherited from the bootstrap project) + +## Unclear Responsibility + +These may be in purview of ha-cluster-bootstrap, or may be in appliance image: + +* install debuginfo packages +* enable coredumps + + +## General / Random + +* csync2_remote assumes there's only one group in csync2.cfg, or, more to + the point, will only add new hosts to the first group. +* Likewise, ssh_merge grabs all hosts regardless of what group they're in + (although this is probably fine) +* get rid of curses junk in log file (fix ENV term) +* Multi-device SBD (use multiple -s args) +* Start new node on standby + # crm configure node node-1 attributes standby="on" + # crm node clearstate node-1 (requires confirmation) + - start new node, unbelievably it works! + # crm node online node-1 +* don't error to log if log not started +* is "partx -a " any sort of sane replacement for partprobe? +* Use ssh-copy-id instead of manual fiddling with authorized_keys? + + +## STONITH Config + +* See https://bugzilla.novell.com/show_bug.cgi?id=722405 for stonith timeout suggestions + + +## Template Mode + +Generally specific to OCFS2 template ATM, as that's the only one extant. + +* Very long path to partition (/dev/disk/by-path/... for iSCSI) means we + can't determine paths to new partitions, thanks to bnc#722959. Unclear + if this will be fixed for SP2. +* /dev/disk/by-id/dm-name paths are unreliable (at least for determining + partitions after carving the device up). +* Probably need to prompt user for new partitions after carving, should + they not be found (FFS). +* ocfs2 template is not the same as Hawk's. Consider enhancing + ha-cluster-bootstrap so it uses Hawk's templates directly rather than using + its own. +* Ensure required RPMs are installed when running template (they're just + Recommends in the spec) +* Specifying sbd without ocfs2 partition may be incompatible with ocfs2 + template (need to test) +* block device size " blockdev --getsz" etc. (when making OCFS2 partition + with "-T vmstore") + -- cgit v1.2.3