Upgrade on-prem Access (Cluster)

[Virtual_Leo]

Hello

We run Access on-premise in a cluster of 3 nodes, version 23.09. We now want to upgrade it to 24.07, when looking at the upgrade documentation it says to remove all nodes except one, and now upgrade the node that still is connected to the loadbalancer.
Upgrade a Workspace ONE Access Cluster (omnissa.com)

This sounds kinda odd to me, is there really no way to upgrade with 0 downtime for the end user?

We have a completly HA Horizon environment with all components loadbalanced, two pods, multiple UAGs, AppVolume Managers, Connectionservers etc.

The part that currently makes the environment not 100% serviceable without any disruption is Access, which is being more and more important in our environment

[Sascha Warno]

Sadly that is one of the main drawbacks of Workspace ONE Access in its current architecture On-Premises.

[Hussam Rabaya]

To ensure a smooth upgrade and a stable service startup, downtime is necessary when upgrading Workspace ONE Access (On-Premises). It is recommended to schedule the upgrade during periods of low user activity. If you have a disaster recovery site, you may consider utilizing it as part of your upgrade strategy. Please note that users connected to Horizon will not be disconnected until they choose to log out of their sessions.

[Virtual_Leo]

Ok thank you for the answers

Is Access with a SaaS subscription expected to have 0 downtime during upgrades?

Edited August 28 by Virtual_Leo

[Sascha Warno]

Correct Access as SaaS service has 0 downtime. New features are constantly checked in and made available during monthly releases.

[Simon Frankiewicz]

Hi, I have a problem updating Access. Maybe someone has encountered this problem.

Installing OS RPMS
Installing Create Repo RPM
Verifying packages...
Preparing packages...
createrepo_c-0.11.1-5.ph3.x86_64
19:26:54: posix_spawn avoided (fd close requested)
19:26:54: posix_spawn avoided (fd close requested)
19:26:54: posix_spawn avoided (fd close requested)
19:26:54: posix_spawn avoided (fd close requested)
Directory walk started
Directory walk done - 252 packages
Temporary output repo path: /opt/vmware/horizon/4.0/.repodata/
Preparing sqlite DBs
Pool started (with 5 workers)
Pool finished
Install photon-upgrade rpm
Verifying packages...
Preparing packages...
photon-upgrade-1.1-1.ph3.noarch
The openssl-c_rehash 1.x package is installed. Removing openssl 1.x rpms before upgrade to photon4
error: package openssl-c_rehash-1.0.2zi-1.ph3.x86_64 is not installed
error: package openssl-1.0.2zi-1.ph3.x86_64 is not installed
2024-09-10T19:26:54+00:00 Upgrading Photon OS from 3.0 to 4.0. Assuming that data backup has already been done.

2024-09-10T19:26:54+00:00 Recording list of all installed RPMs on this machine to /tmp/photon-upgrade.sh-2024-09-10-19-26-54/rpm-qa.txt.
2024-09-10T19:26:55+00:00 Creating backup of RPM DB.
2024-09-10T19:26:55+00:00 Generating tdnf cache on Photon OS 3.0
Error(1238) : Couldn't read a file:// file
Error: Failed to synchronize cache for repo 'vCenter bundled Photon OS 3.0 (x86_64)' from 'file:///opt/vmware/horizon/3.0'
Refreshing metadata for: 'vCenter bundled Photon OS 3.0 (x86_64)'
Disabling Repo: 'vCenter bundled Photon OS 3.0 (x86_64)'
Metadata cache created.
2024-09-10T19:26:55+00:00 The following deprecated packages will be removed before upgrade -
    libxml2-python
    python-lxml
    python-six
    python-xml
    python2
    python2-libs
2024-09-10T19:26:55+00:00 Generating tdnf cache on Photon OS 4.0
Refreshing metadata for: 'vCenter bundled Photon OS 4.0 (x86_64)'
Metadata cache created.
2024-09-10T19:26:55+00:00 Finding packages in the target repo. This may take several minutes.
2024-09-10T19:26:56+00:00 Error: Packages - libdnet tcpdump - are not available in the target repo for OS upgrade. Please provide the correct repo having the required packages.
Original list of RPMs and RPM DB are stored in /tmp/photon-upgrade.sh-2024-09-10-19-26-54,  please provide contents of that folder along with system journal   logs for analysis; these logs can be captured using command-
 # /usr/bin/journalctl -xa > /tmp/photon-upgrade.sh-2024-09-10-19-26-54/journal.log
 Cannot continue. Aborting.
OS upgrade from photon 3 to photon 4 failed.
 

 

 

 

 

[Sascha Warno]

I guess you installed tcpdump on the old nodes before for troubleshooting, as this is an offline upgrade and also upgrades the OS to version 4 of PhotonOS and we did not include the version of tcpdump and libdnet for that it fails. Remove tcpdump and libdnet on the nodes and try again.

rpm -e libdnet 
rpm -e tcpdump

and then try the upgrade again.

[Simon Frankiewicz]

Thanks,
I'll try to do that.

[Virtual_Leo]

On 9/16/2024 at 1:15 PM, Simon Frankiewicz said:

Thanks,
I'll try to do that.

If you haven't already tried that should solve it.

I had to expand the disk to be able to upgrade, installed parted according to photon documenation (Expanding Disk Partition · VMware Photon OS 3.0 Documentation) and ran the update, it failed after a while because of this package. Luckily i was able to run it again with not too much service interuption. 

I cannot find the Photon OS upgrade being mentioned anywhere, atleast that would have been nice to know.

Clearly focus is on the cloud product of Access. This ofcourse sucks for on-premise users that has extremely strict security policies and can't move to the cloud.

[Simon Frankiewicz]

Thanks for the tip, uninstalling the packages solved the issue.

[Noordan]

Hello, We are trying to upgrade a Access 23.09 Appliance (single node cluster), but the upgrade fails/hangs after a while. We waited for it to continue, but nothing more happend on the appliance.

In the update log we got two errors, and I dont know if they are related to this. Please see below log output. 

Anyone that have a clue whats wrong here? 

We removed tcpdump and parted with rpm -e before the upgrade. 

Installing OS RPMS
Installing Create Repo RPM
error: Failed dependencies:
	libcrypto.so.1.0.0()(64bit) is needed by createrepo_c-0.11.1-5.ph3.x86_64
	libssl.so.1.0.0()(64bit) is needed by createrepo_c-0.11.1-5.ph3.x86_64
04:56:11: posix_spawn avoided (fd close requested) 
04:56:11: posix_spawn avoided (fd close requested) 
04:56:11: posix_spawn avoided (fd close requested) 
04:56:11: posix_spawn avoided (fd close requested) 
Directory walk started
Directory walk done - 252 packages
Temporary output repo path: /opt/vmware/horizon/4.0/.repodata/
Preparing sqlite DBs
Pool started (with 5 workers)
Pool finished
Install photon-upgrade rpm
Verifying packages...
Preparing packages...
photon-upgrade-1.1-1.ph3.noarch
The openssl-c_rehash 1.x package is installed. Removing openssl 1.x rpms before upgrade to photon4
error: package openssl-c_rehash-1.0.2zi-1.ph3.x86_64 is not installed
error: package openssl-1.0.2zi-1.ph3.x86_64 is not installed
2024-10-18T04:56:11+00:00 Upgrading Photon OS from 3.0 to 4.0. Assuming that data backup has already been done.

 

[Sascha Warno]

there should be more output to the process. can you check for the end  which should usually be,

Cannot continue. Aborting.

OS upgrade from photon 3 to photon 4 failed.

Just before that should be the reason why it cannot proceed, usually because a package is installed it can't upgrade and you need to remove before upgrade.

 

[Noordan]

We did not recivie a "failed" result. This is the end of the file. 
Did we just wait to short in that case? (we waited 10-15 minutes but the upgrade did not continue)

We also have a three node cluster that we have upgraded successfully, and in that case the upgrade continute directly with the next steps in the upgrade.