haproxytech
OVA doesn't get configured (e.g. no IP) with vSphere 8.0 U3
Not so confident but I think this OVA doesn't work properly on later releases of vSphere (atm 8.0 u3)
It seems there is a problem with how it still (last update 2021 ... ) uses the VMwareGuestInfo cloud-init datasource, which I believe is deprecated.
The mechanism I have figured out is that this OVA first executes /var/lib/vmware/ovf-to-cloud-init.sh which is responsible of creating the userdata/metadata that is fed to cloud-init in the later phase (handled by systemd)
This script will evoke, ovf-rpctool set userdata "$encoded_userdata" and ovf-rpctool set userdata.encoding "base64" which I guess its suppose to configure the guestinfo.userdata and metadata.
However during startup cloud-init doesn't find that userdata, and thus fails to configure anything.
/var/log/cloud-init.log shows the following.
2024-11-14 20:57:18,725 - util.py[DEBUG]: Running command ['/usr/bin/vmware-rpctool', 'info-get guestinfo.metadata'] with allowed return codes [0] (shell=False, capture=True)
2024-11-14 20:57:18,729 - DataSourceVMwareGuestInfo.py[DEBUG]: No value found for key metadata
2024-11-14 20:57:18,729 - DataSourceVMwareGuestInfo.py[DEBUG]: loaded metadata {}
2024-11-14 20:57:18,729 - DataSourceVMwareGuestInfo.py[DEBUG]: Getting guestinfo value for key userdata
2024-11-14 20:57:18,729 - util.py[DEBUG]: Running command ['/usr/bin/vmware-rpctool', 'info-get guestinfo.userdata'] with allowed return codes [0] (shell=False, capture=True)
2024-11-14 20:57:18,731 - DataSourceVMwareGuestInfo.py[DEBUG]: No value found for key userdata
2024-11-14 20:57:18,731 - DataSourceVMwareGuestInfo.py[DEBUG]: Getting guestinfo value for key vendordata
2024-11-14 20:57:18,731 - util.py[DEBUG]: Running command ['/usr/bin/vmware-rpctool', 'info-get guestinfo.vendordata'] with allowed return codes [0] (shell=False, capture=True)
2024-11-14 20:57:18,733 - DataSourceVMwareGuestInfo.py[DEBUG]: No value found for key vendordata
Btw I get this permission error so it may be due to how no longer later vsphere versions don't allow to directly update this value.
root@localhost [ ~ ]# vmware-rpctool "info-set guestinfo.userdata aaa"
Permission denied
Not elegant, but the current only workaround I found
[ Current workaround ]
- Deploy this OVA and start as guided and expect to fail (no IP, and no password configured)
- Force into shell using https://vmware.github.io/photon/assets/files/html/3.0/photon_troubleshoot/resetting-a-lost-root-password.html. Somehow get the IP address configured with sshd started from there.
- Save contents of
/var/lib/vmware/encoded_metadata.txt,/var/lib/vmware/encoded_userdata.txt - Destroy VM
- Redeploy OVA with the exact same values.
- Before starting the VM, check each mac address and match it to the values in
/var/lib/vmware/encoded_metadata.txt(by decoding it) - Before starting add the following from Advanced Config in the edit VM option. Use the save contents on the previous procedure.
Hi. I'm working right now with vSphere 8.0.3 with Tanzu and using HAProxy with this OVA... I think this OVA is no maintained too... but I've managed to upgrade to Photon OS 5.0, HAProxy 2.8.2 and dataplaneapi 2.7.1... all is working perfectly with vsphere v.8.0.3 with Tanzu. I have some PowerShell code to do it automatically, but it's simple:
- You need to upgrade all packages in Photon OS 3.0 with "tdnf upgrade --nogpgcheck -y"
- You need install Photon upgrade script: "tdnf -y install photon-upgrade"
- Then upgrade to Photon 4.0: "photon-upgrade.sh --upgrade-os --assume-yes" and then reboot.
- Then upgrade again to Photon 5.0: "photon-upgrade.sh --upgrade-os --assume-yes" and then reboot.
- Your HAProxy and "dataplaneapi" will be upgraded to 2.8.2 and 2.71, but they will not work with your actual configuration.
- HAProxy new configuration file doesn't work. It needs to be rewrited with "/etc/haproxy/haproxy.cfg.lkg" that it's the last backup from initial configuration.
- "dataplaneapi" new version will not be in "/usr/local/bin", where is in the original OVA, but in "/usr/libexec/haproxy/", you can write the new one over the old one or modify the "/etc/systemd/system/dataplaneapi.service" file to point to the new one.
- The dataplaneapi.service need to know how to "restart" HAProxy service (in old version not to)
- You need to add line "RESTART_CMD=/usr/bin/systemctl restart haproxy" to "/etc/haproxy/dataplaneapi.cfg"
- Then to the service in "/etc/systemd/system/dataplaneapi.service" file, add a line too: "--restart-cmd=${RESTART_CMD}"
- There is another problem in the four services that do things with HAProxy service, that I don't know yet what is the cause, but I know how to solve. The need to add a line "DefaultDependencies=no" in the "[Unit]" Section, because, if not, they will not start on boot. Those services files are:
- /etc/systemd/system/anyip-routes.service
- /etc/systemd/system/route-tables.service
- /etc/systemd/system/dataplaneapi.service
- /lib/systemd/system/haproxy.service
With those modifications your HAProxy will work with Tanzu.
There is also a Broadcom KB regarding this issue. Is a much simpler workaround which worked for me. https://knowledge.broadcom.com/external/article/380676/haproxy-vm-deployed-on-esxi-80-u3b-fails.html
@vrhinow This is the same fix that has been already merged into this repo: https://github.com/haproxytech/vmware-haproxy/pull/34
@akutz Can you publish a new release of the current state of this repo? Your fix is urgently needed :)
For people using Ansible, deploy the OVA as usual, but powered-off, then fix it as follows:
- name: Fix HAProxy
community.vmware.vmware_guest:
advanced_settings:
- key: guest_rpc.auth.cloud-init.set
value: 'FALSE'
cluster: MyCluster
datacenter: MyDatacenter
folder: /MyDatacenter/vm/MyFolder
name: HAProxy
- name: Boot HAProxy for the first time
community.vmware.vmware_guest:
cluster: MyCluster
datacenter: MyDatacenter
folder: /MyDatacenter/vm/MyFolder
name: HAProxy
state: poweredon
wait_for_ip_address: true
