Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

vCenter HA Degraded: 'Appliance configuration is out of sync' or 'PostgreSQL replication is not in progress'

Technical article
VMware

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

When vCenter High Availability (VCHA) shows Degraded, one of the three nodes is not doing its job and automatic failover is not available. Two messages account for most of the reports: "Appliance configuration is out of sync." and "PostgreSQL replication is not in progress." Broadcom documents several unrelated causes for them. The most common are SSH between the nodes not working, a stray root-owned file in the PostgreSQL directory, and an unreliable replication network.

Do not delete and redeploy the cluster as a first move. Read the VCHA log on the node that is complaining, match the first error line to the table below, and fix that one cause.

What the error means

Per the vSphere 7.0 documentation (Last Updated November 4, 2026), a healthy cluster needs the Active, Passive and Witness nodes all fully operational. If any node fails, the cluster is degraded and failover cannot occur. The page says the state appears as events, alarms and SNMP traps, and that for replication failures you should check the vCenter HA network has enough bandwidth and keep latency at 10 ms or less.

The two messages are symptoms, not diagnoses:

  • Out of sync means the Active node could not copy its configuration to the Passive node. KB 432952 explains that the sync uses rsync, which needs SSH. SSH can be turned off after the first sync for security reasons, but then the cluster falls out of sync whenever a resync is needed, for example after a failover or a configuration change.
  • PostgreSQL replication is not in progress means the Passive node's database copy is not following the Active node's. KB 383199 lists service interruptions on the Passive node and network problems on the replication interfaces as the causes, and says to confirm the PostgreSQL server is running on the Passive node and reachable on the vCenter HA network.

Likely causes

Cause What you see How to confirm
SSH disabled or sshd stopped on a node "Appliance configuration is out of sync.", state Degraded, Enable greyed out after Maintenance Mode (KB 432952); after failover, port 22 refused on the old Active node (KB 418705) /var/log/vmware/vcha/vcha.log shows SSH connection refused to the peer; check the sshd service
SSH key authentication between nodes fails Configuring VCHA fails with "Appliance state is out of sync." (KB 372362) vcha.log shows rsync exit code 255 and "Connection closed by" on port 22; sshConnect.log shows "Unable to connect to port 22"
Root-owned file in the PostgreSQL directory "PostgreSQL replication is not in progress." on enabling VCHA; storage partition on the Passive node keeps growing (KB 438588) repl_passive_setup.log on the Passive node shows could not open file "./postgresql.orig": Permission denied
Replication network wrong or slow Degraded with the replication message (KB 383199) The two interfaces are on different port groups, outside the subnet, or latency is above 10 ms
Heartbeat network drops packets Nodes keep swapping roles; vCenter name intermittently unreachable (KB 430364) vcha.log contains "Slave timed out", "Lost master" or "Startup Timeout"

Prove the cause before changing anything

All of these are read-only. These commands have not been run or lab-tested.

  1. Read the state. In the vSphere Client, select the vCenter Server and open Configure > vCenter HA. The page shows the cluster state and the message. Broadcom documents this UI as the place the replication error appears (KB 383199); Broadcom documents no command-line status check for vCenter HA.
  2. Read the VCHA logs. On the node that reports the problem, the logs live in /var/log/vmware/vcha/. KB 372362 names vcha.log and sshConnect.log. KB 438588 names repl_passive_setup.log, on the Passive node.
  3. Look for these lines. A bad result for an SSH fault is a "System command failed" entry with exit code 255 and the rsync text "Connection closed by" the peer on port 22, and repeated "Unable to connect to port 22" lines. A good result is no such lines in the recent log. For the file-permission fault the bad line is could not open file "./postgresql.orig": Permission denied, repeating on each base backup attempt.
  4. Check ownership of the database directory. On the Active node (KB 438588):
cd /storage/db/vpostgres
ls -la

Good: files are owned by vpostgres:vpgmongrp. Bad: a file owned by root:root. The KB says these are often manual backups such as postgresql.orig or postgresql.backup.

  1. Check the network. KB 430364 gives these checks from the VCSA shell of the Active node, with <Peer_IP> the heartbeat address of the Passive or Witness node:
vmkping -I eth1 <Peer_IP>
ping <Peer_IP> -c 100

The KB expects round-trip time consistently under 10 ms and says any packet loss can trigger a failover. It prints vmkping itself; if vmkping is not available in your appliance shell, use the plain ping.

Example (illustrative): vc01 shows VCHA Degraded after a patch window. vcha.log on the Active node shows rsync failing with exit code 255 and "Connection closed by" the Passive node. SSH had been turned off on the Passive appliance after the first deployment. This matches KB 432952, not a database problem, so there is no reason to touch PostgreSQL.

Fix

SSH disabled or stopped

If SSH was turned off, enable it (KB 432952). If sshd is stopped or masked on the previously Active node after a failover, KB 418705 gives:

systemctl restart sshd
systemctl unmask sshd.service
systemctl restart sshd

Run the unmask line only if the first restart fails because the service is masked. Then retry the failover from the vSphere Client.

Disruptive step: failing over vCenter HA briefly interrupts vCenter. Do it outside busy periods and make sure the Witness node is up first. If you prefer to avoid failover, put vCenter HA into Maintenance Mode (below) while you repair SSH.

SSH key authentication fails

KB 372362 (vCenter 8.0) gives these steps: generate a key pair for root with ssh-keygen; append the public key to /root/.ssh/authorized_keys (the KB's example is cat /root/.ssh/id_ed25519.pub >> /root/.ssh/authorized_keys, matching whichever algorithm you generated); set PreferredAuthentications publickey,password,keyboard-interactive,gssapi-with-mic in /etc/ssh/sshd_config; disable VCHA; reboot the appliance; re-enable VCHA.

The KB's additional notes mention a different method that conflicts with its main steps; follow the main steps.

Disruptive step: disabling VCHA removes the HA protection and the reboot interrupts vCenter. Schedule this. To back out, re-enable VCHA; there is no undo for the reboot itself.

Root-owned file in the database directory

From KB 438588: on the Active node, find the root-owned file with ls -la in /storage/db/vpostgres, move or delete it, then restart the Passive node. The KB gives no exact commands for the move or restart.

Risk: this is a live vCenter database directory. Move the file elsewhere rather than deleting it, and confirm the file is a manual copy and not part of the running database before you touch it.

Replication network problem

KB 383199 lists, in order: both replication interfaces on the same port group; addresses in that port group's subnet or VLAN; latency between Active, Passive and Witness under 10 ms; reboot the Passive node to restore its services; and, if the problem persists, remove the vCenter HA configuration and redeploy it.

Disruptive step: removing and redeploying vCenter HA is the last resort. Plan for a period without HA protection.

Nodes flapping between Active and Passive

KB 430364 attributes this to packet loss or high latency on the dedicated heartbeat network. It says the VCHA network interfaces (usually eth1) should be on a dedicated, isolated VLAN, MTU must match across virtual and physical switches, and nothing should delay heartbeat traffic.

If you cannot fix the network immediately, the KB's workaround is Configure > vCenter HA > Edit > Maintenance Mode. Replication continues but automatic failover stops. Risk: with automatic failover off, an Active node failure needs a manual response.

Failover failed and the Passive node is not serving

The 8.0 failover documentation (Last Updated October 28, 2026) gives the causes as the Witness being unavailable while the Passive node tries to take over, or a state synchronisation problem. If the Witness recovers, log in to the Passive node console, enter shell, run vcha-reset-primary and reboot it. If neither Active nor Witness can be recovered, delete their VMs, then enter shell and run vcha-destroy on the Passive node and reboot.

Disruptive step: vcha-destroy is the step Broadcom gives to force the Passive node to become a standalone vCenter. Broadcom's page does not explain what it removes; read the vCenter HA documentation before using it.

Confirm the fix

  • Configure > vCenter HA shows the cluster as healthy, with all three nodes up, and no replication or sync message.
  • vcha.log no longer shows "System command failed", "Unable to connect to port 22" or "Slave timed out".
  • ls -la in /storage/db/vpostgres shows no root-owned stray files.
  • If you used Maintenance Mode as a workaround, turn it off so that automatic failover is restored.

What to watch afterwards

  • After any failover or vCenter HA configuration change, check the cluster state within a few minutes.
  • Keep the SSH service running on all three nodes if you rely on a resync.
  • Watch for latency over 10 ms between nodes, and for packet loss, on the vCenter HA network.
  • Do not leave manual copies in /storage/db/vpostgres.

Risks and limits

  • The degraded-cluster behaviour is described from the vCenter 7.0 documentation; the KB pages cover 7.x and 8.x as listed in the version scope.
  • vCenter HA has no documented command-line status check; use the vSphere Client page and the logs.
  • KB 372362 contains two different methods for the SSH key; follow the main steps only, and test on your build.
  • vmkping is printed in KB 430364 for a VCSA shell. It normally belongs to ESXi, so ping is the safe check.

NetLeverage runs its own hosting on VMware vSphere, vCenter and ESXi. Talk to an engineer.

Sources