Forum
Welcome, Guest
Username: Password: Remember me
  • Page:
  • 1
  • 2

TOPIC:

HA deliverables update window 10 years 3 months ago #167

Hi,

As you may have noticed, Citrix released yesterday the ESP1 and some other updates to XenServer 6.2 affecting the kernel.

My question is if you have an estimate between those releases and the update of the packages on your side for those kernel versions to avoid compiling them myself.

Also, even if the upgrade/¿update too? is disencouraged
"AVOID Upgrading XenServer after completing this how-to – installed packages and configurations will
likely be lost as part of the upgrade process (eg. Upgrading from version 6.1 to 6.2)"

Is it possible to upgrade/update and reinstall the packages in any method that leaves the pool in an stable status?

BR

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

HA deliverables update window 10 years 3 months ago #170

Thanks for posting this. We generally can have updated modules within a couple of days. A patch release for iscsi-ha should be ready and posted by Friday which will supply the new kernel module.

FYI - Kernel updates only affect the iscsi-ha module which depends on DRBD for replication.

Regarding the documentation which discourages major updates - this is noted simply because the installation of ha-lizard/iscsi-ha will be erased as part of the XenServer upgrade process. In this case it can be backed up and then reinstalled after the server upgrde is complete. For now - one would have to complete the installation steps again.

We are working on a new installer that will simplify installation in the future. The goal is to run the installer, answer a few questions and then have a working 2-node HA pool within minutes. More to come on this...

Lastly - regarding upgrading and leaving the pool in a stable status - if only running ha-lizard - then there are no issues with stability. If running iscsi-ha for replication - one should use caution as the disks should only be exposed to XenServer through DRBD which would experience some downtime durring a major release update.
The following user(s) said Thank You: Antonio

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

HA deliverables update window 10 years 3 months ago #171

Thanks for the reply.

It seems a reasonable window to wait for official releases after the patches are published by citrix. And yes, I'm specially concerned about iscsi-ha.

Following the install process it looks like the upgrade/update of any component of XenServer followed with an update/reinstall of iscsi-ha/ha-lizard should be possible, I just wanted to make sure no uuids or something similar are used which will be lost during the update-reinstall process letting the system unusable/unstable. A reasonable downtime when a mayor upgrade have to be performed can be assumed.

Regarding the installer, it's great news. Currently i'm implementing a method to perform the install/update from an ansible playbook, but the installer would make it less error prone an much more easy to perform those operations.

Thnaks again for you answer and you work!

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

HA deliverables update window 10 years 3 months ago #172

No UUIDs are stored statefully by halizard. All the UUIDs (SRs, VMs, hosts, etc..) are dynamically read and processed in real time. So, it is safe to have changes in the pool without affecting halizard.

There are, however, 2 situations where this is not true:

1- fencing configuration files (HP ILO or custom scripts). HAlizard passes the UUID of the host to be fenced to the configured fencing method. Any fencig configuration files would need to be updated if the UUID of a host changes.

2- If operating in mode 1 (managing appliances), there is an option to exlcude certain appliances. The excludes are declared in the configuration by UUID of the appliance.. This is a rarely used feature though.

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

HA deliverables update window 10 years 3 months ago #173

Antonio
an iscsi-ha release candidate is available for download - version 1.7.6. You can either try the new release - or, use the prebuilt DRBD RPMs supporting the latest XenServer kernel (SP1).

iscsi-ha release at the following link:
www.halizard.org/release/iscsi-ha/

DRBD Packages for all kernels since XenServer 6.1 and XCP 1.6
www.halizard.org/release/drbd/RPM/

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

HA deliverables update window 10 years 2 months ago #175

We've documented a procedure for dealing with a rolling update in a 2-node pool with HA-Lizard+iSCSI-HA. There is no downtime when following these steps. Below are the steps. These have also been incorporated into the latest how-to documentation which is available for download here:

halizard.org/release/doc/how-to/iscsi-ha...owto_1.3.7_final.pdf

Ha-Lizard/iSCSI-HA 2-node pool: Procedure for XenServer rolling update where Master is updated first

!! IMPORTANT NOTES AND CONSIDERATIONS !!
** This procedure is only required if a XenServer patch requires a reboot AND the kernel version is changing to a newer kernel

** Below steps assume device name is xenbr0 and iscsi IP is 10.10.10.3 and DRBD device is iscsi1- replace with your values!!

** If the XenServer upgrade includes an kernel update - you must have the matching DRBD kernel module available. It should be included in the latest

release of iscsi-ha. Before proceeding make sure you have a DRBD kernel module that exactly matches your new XenServer kernel version.

Before proceeding, you should have the following values handy:
- the network device name used by iscsi-ha (usually xapi0 or xenbr0)
- IP address used by iscsi-ha to expose the iscsi target (usually 10.10.10.3)
- DRBD device name (usually iscsi1 if built by following the how-to)

Upgrade Steps:

1 - Migrate all VMs to the slave host

2 - Disable HA-Lizard "ha-cfg status" then "yes" to disable. Wait about 1 minute. This only needs to be done on any one host in the pool

3 - Stop iSCSI-HA on both hosts and make sure it stays off until the upgade is complete - on both hosts
"service iscsi-ha stop -w"
"chkconfig iscsi-ha off"
"chkconfig iscsi-ha-watchdog off"

** The following 2 steps should be performed quickly so that disk access is only lost for a few seconds

4 - Stop exposing disks through master - perform this step on the Master only
"service tgtd forcedstop && rm -f /var/lock/subsys/tgtd && drbdadm secondary iscsi1 && ip addr del 10.10.10.3/24 dev xenbr1"

5 - Start exposing disks on slave - perform this step on the Slave only
"drbdadm primary iscsi1 && service tgtd start && ip addr add 10.10.10.3/24 dev xenbr1"
"arping -I xenbr1 -U 10.10.10.3 -c 2 -w 2"

At this point all your VMs should be running on the slave with full access to their disks.

6 - Perform the upgrade on the Master per the Citrix instructions and reboot

7 - With the master upgraded and rebooted - install a new DRBD kernel module on the master.
" cd /etc/iscsi-ha/RPM/`uname -r` "
" rpm -ivh drbd-km-2*.rpm "

8 - Stop exposing disks on the slave - perform this step on the Slave only
"service tgtd forcedstop && rm -f /var/lock/subsys/tgtd && drbdadm secondary iscsi1 && ip addr del 10.10.10.3/24 dev xenbr1"

9 - Start exposing the disks on the Master and start services - perform this step on the Master only
" chkconfig iscsi-ha on "
"chkconfig iscsi-ha-watchdog on"
" service iscsi-ha start -w "

10 - Migrate all VMs to the Master

11 - Perform the upgrade on the slave per the Citrix instructions and reboot

12 - With the Slave upgraded and rebooted, install a new DRBD kernel module on the slave.
" cd /etc/iscsi-ha/RPM/`uname -r` "
" rpm -ivh drbd-km-2*.rpm "

13 - Start services on the slave
" chkconfig iscsi-ha on "
"chkconfig iscsi-ha-watchdog on"
" service iscsi-ha start -w "

14 - Finally - enable HA. Only required on any one of the two hosts
"ha-cfg status" then "yes"
The following user(s) said Thank You: Antonio

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

Last edit: by Pulse Supply. Reason: mark as resolved
  • Page:
  • 1
  • 2