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.
The vSphere Client page that says only "no healthy upstream" is, in Broadcom KB 390105's words, "an alternate error message for HTTP Return code 503". Several causes are documented: a stopped service, a full filesystem, an expired certificate, a damaged certificate store, a wrong clock, or a configuration file that stops the vpxd service from starting. Work through the checks below to tell them apart before you restart anything.
What the error means
Broadcom describes the message as a 503: the server is "temporarily unable to handle the request" (KB 390105). Broadcom's KBs do not describe the proxy design. KB 437959 names the rhttpproxy and envoy services as the pair whose logs show the fault after a vCenter 7.0 reboot. In plain terms, a front-end proxy on the vCenter Server Appliance (VCSA) passes each request to the service that owns the page, such as the sign-in service or the vSphere Client service, and has nothing to pass it to when that service is stopped or refusing connections.
KB 390105 (Updated On: 10-07-2026, vCenter Server 7.x and 8.x) lists these causes of the vpxd service not running or not accepting connections: expired certificates, disk space issues, "Running out of sessions (reached the maximum limit of 2000)", the service running out of memory, and scheduled maintenance such as updates. It also lists the SSO service (vmware-stsd) being down, a modified /etc/hosts that prevents services from starting, the lookup service lookupsvc failing to start on an LDAP client error, and incorrect date and time on the vCenter.
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| Services stopped after a reboot (vCenter 7.0 only) | Message persists "even though sufficient time has passed since the reboot" | rhttpproxy and envoy logs show "unknown cluster" errors (KB 437959; the KB says the issue is resolved in 8.0) |
| Expired Machine SSL or STS certificate | Message plus errors such as "certificate has expired"; VAMI sign-in may fail | service-control --status --all shows vmware-stsd, vmware-vpxd or vsphere-ui stopped; certificate status tool (KB 413685, 418844) |
| Appliance filesystem full | Services exit or will not start | df -h shows / or /storage/log at or near 100%; vmon.log shows "No space left on device" (KB 410218, 378472) |
| Certificate store (VECS) damaged | Services will not start even with service-control --start --all |
vecs-cli store list shows only a few stores; vmafdd.log shows "ERROR! [VecsIpcGetEntryByAlias] is returning [4312]" (KB 423397) |
Damaged vpxd.cfg |
Starting vmware-vpxd fails; core dumps in /var/core |
The vpxd start error, and /var/run/log/vpxd/vpxd.log not updating with new entries (KB 425518) |
| Datastore under the appliance full | Appliance boots into emergency mode | VM console shows the file system check failure; datastore free space in the Host Client (KB 405611) |
Wrong date and time, LDAP error, edited /etc/hosts |
Services fail on start | KB 390105 lists each as a cause; check the clock and the file by eye |
Prove the cause before changing anything
Everything in this section only reads. Use SSH to the appliance as root; the KBs use the appliance shell. KB 425518 shows shell being entered after the SSH login before cp and vi are used, so run shell before using cp, find or an editor.
1. Read the service list. KB 413685 shows this check:
service-control --status --all
The output has a "Running:" list and a "Stopped:" list. A good result has vmware-stsd, vmware-vpxd and vsphere-ui under Running. A bad result has one of them under Stopped, as in the failing cases of KB 413685 and KB 423397. Several other services also appear under Stopped in KB 413685's sample, so judge by the three named here and by whichever service the logs point to.
2. Read the disk.
df -h
Look at the Use% column for / and for /storage/log. KB 410218 applies 95% to /storage/log: it tells the reader to locate the partition for /storage/log and, if Use% is 95% or higher, to clear space. KB 410218 shows services exiting with "Exit code 120" when vmon cannot write its PID files because the partition is full.
3. Read the service manager log. /var/log/vmware/vmon/vmon.log records service start and exit lines (KB 413685 and 410218). "No space left on device" points to step 2. A certificate error such as "certificate has expired" points to the certificate cause.
4. Check the reverse proxy logs after a reboot. /var/log/vmware/rhttpproxy/rhttpproxy-##.log and /var/log/vmware/envoy/envoy-##.log. KB 437959 shows lines containing "unknown cluster" in the vCenter 7.0 reboot case.
5. Check the certificate store. In the failing case of KB 423397, the command below listed only MACHINE_SSL_CERT, TRUSTED_ROOTS and TRUSTED_ROOT_CRLS. The KB lists "Unable to list solution user certificates in vecs" among its symptoms, and its vmafdd.log line reads "ERROR! [VecsIpcGetEntryByAlias] is returning [4312]":
/usr/lib/vmware-vmafd/bin/vecs-cli store list
6. Check the clock against a trusted clock. KB 390105 lists wrong date and time as a cause.
Fix
Before any change below: take a snapshot of the vCenter VM. KB 390105 says to do so before making any changes, and for vCenters in Enhanced Linked Mode it points to its own snapshot procedure. Reverting a snapshot discards everything done after it was taken; where a step below has its own back-out, use that first.
Services stopped after a reboot (vCenter 7.0)
Disruptive step: the workaround below stops every vCenter service, so the vSphere Client and anything else that talks to vCenter is unavailable until the services return. Save first: take the snapshot above, and run service-control --status --all and keep its output, so you can compare the Running and Stopped lists afterwards. Back out: there is nothing to undo; if services do not return, read vmon.log and compare with the saved list before trying again.
KB 437959 gives this workaround, and says the issue is resolved in vCenter Server 8.0:
service-control --stop --all && service-control --start --all
Full filesystem
A full appliance partition is covered by KB 378472 and KB 410218. For finding and clearing it, see "vCenter appliance partition full: finding what filled /storage/log, /storage/db or /storage/seat" (vcenter-appliance-partition-full-log-database). Risk: deleted files cannot be recovered except by reverting the snapshot, so do not remove anything under /storage/db or another data volume. Disruptive step: after space is free, restart services with the command in the previous section; the vSphere Client is unavailable while they restart. Then run df -h again.
Datastore under the appliance full
KB 405611 describes a different case: the datastore under the appliance is full, and the appliance boots into emergency mode. Its first step is to power off the vCenter VM, expand the datastore, and confirm the datastore shows free space.
Disruptive step: powering off the vCenter VM makes vCenter unavailable until it is powered on again. Save first: note the datastore's current capacity and free space in the Host Client (Storage, select the datastore), and take the snapshot the KB asks for before any file-system repair. Back out: a datastore expansion is not shrunk by any step here, so agree the new size with your storage team first; power the VM back on once the datastore shows free space.
Expired certificates
KB 413685 and KB 418844 point to the vCert script in KB 385107. To identify which certificate expired and renew it, see "vCenter Sign-In Fails After a Certificate Expires: Find the Store Responsible" (vcenter-sign-in-fails-expired-sts-machine-certificate). KB 385107 states the script "is intended to be used at the direction of Broadcom Global Support" and warns that "Changes made could render this system inoperable".
Risk: a certificate replacement affects every service and every solution that trusts vCenter. Save first: KB 385107 says to ensure you have a valid VAMI-based backup or offline snapshots of all vCenter/PSC nodes in the SSO domain, so take those rather than relying on the snapshot above alone. Back out: restore from that backup or revert the offline snapshots, only if no other changes have been made since.
Damaged certificate store or configuration file
For a damaged VECS store, KB 423397 gives one resolution: "Restore vCenter Server from backup".
Risk: a restore replaces the appliance's state with the state in the backup, so changes made after the backup are lost. Save first: take a snapshot of the current appliance if it can still be taken, and note the backup's date. Back out: revert to that snapshot, if it exists and nothing else has changed.
For a vpxd.cfg with a duplicate enabled tag inside vcls, KB 425518 starts with a copy of the file (the KB names the copy vpxd.cfg.bak; a dated name keeps an earlier copy from being overwritten):
cp /etc/vmware-vpx/vpxd.cfg /etc/vmware-vpx/vpxd.cfg.20261120
After the copy, the KB removes the duplicate line in an editor and restarts all services with the command above. Risk: an editing mistake in vpxd.cfg keeps vpxd from starting. Back out: run cp /etc/vmware-vpx/vpxd.cfg.20261120 /etc/vmware-vpx/vpxd.cfg, then restart the services again with the command above. KB 425518 also shows that running /usr/sbin/vpxd reports "Failed reading /etc/vmware-vpx/vpxd.cfg" with "Duplicate child: enabled" while the file is damaged.
Example (illustrative): vcsa01 is restarted after a maintenance window and the vSphere Client shows "no healthy upstream" an hour later. service-control --status --all lists vmware-stsd and vmware-vpxd under Stopped. df -h shows /storage/log at 98%, and vmon.log has "No space left on device". After a snapshot, the administrator clears space by following the partition-full article, restarts all services, and sees df -h fall to 61%. The sign-in page loads.
Confirm the fix
service-control --status --alllistsvmware-stsd,vmware-vpxdandvsphere-uiunder Running.df -hshows enough free space on every partition you cleared.- Sign in to the vSphere Client and open a host, a VM and the task list.
- Wait through one full sign-in cycle (log out and in) to check that sessions are issued.
What to watch afterwards
- Alert on
/and/storage/logusage well before 95%. - Alert on certificate expiry dates, including the STS signing certificate.
- Keep the vCenter time source correct.
- After reboots of vCenter 7.0, check the service list and not only the login page.
Read next
Related articles you may want next.
- VMwarevSphere Client '503 Service Unavailable': find which vCenter service did not startvSphere Client shows 503 Service Unavailable? Run service-control --status --all, read vpxd and vsphere-client logs, then start services in the right way.Read article
- VMwarevCenter appliance partition full: finding what filled /storage/log, /storage/db or /storage/seatvCenter slow, 503 or services stopped with a full /storage/log or /storage/seat? Read-only checks, known causes and the supported disk-growth path.Read article
- VMwarevCenter Sign-In Fails After a Certificate Expires: Find the Store ResponsiblevCenter sign-in fails or services won't start? Identify whether the STS signing, Machine SSL or solution-user certificate expired, then renew it safely.Read article
Risks and limits
- These commands have not been run or lab-tested.
- The Broadcom KBs are written for the VCSA on 7.x and 8.x. A Windows-based vCenter, vCenter 6.x and vCenter 9.0 are not covered here.
- Certificate replacement is the riskiest change here. Ask Broadcom Support first if you are unsure which certificate is expired.
- Other causes exist. KB 390105 says the error can have several reasons.
Sources
- no healthy upstream error when attempting to access the vCenter Server vSphere Client — Broadcom Knowledge Base 390105, Updated On: 10-07-2026, accessed 20 November 2026
- vSphere Client reports no healthy upstream due to expired certificates — Broadcom Knowledge Base 413685, Updated On: 07-24-2026, accessed 20 November 2026
- Unable to launch the vSphere Client. No healthy Upstream — Broadcom Knowledge Base 418844, Updated On: 02-06-2026, accessed 20 November 2026
- no healthy upstream error when attempting to access the vCenter Server with vSphere Client — Broadcom Knowledge Base 423397, Updated On: 12-22-2025, accessed 20 November 2026
- no healthy upstream error in vSphere Client and multiple service startup failures after rebooting vCenter Server 7.0 — Broadcom Knowledge Base 437959, Updated On: 05-26-2026, accessed 20 November 2026
- no healthy upstream error when accessing the vCenter Server UI due to vpxd service crashing — Broadcom Knowledge Base 425518, Updated On: 01-15-2026, accessed 20 November 2026
- VMware Secure Token Service (STS) fails with No healthy upstream error — Broadcom Knowledge Base 410218, Updated On: 04-29-2026, accessed 20 November 2026
- vCenter appliance root partition "/dev/mapper/vg_root_0-lv_root_0" is full — Broadcom Knowledge Base 378472, Updated On: 12-19-2025, accessed 20 November 2026
- No Healthy Upstream Error When Accessing vSphere Client in vCenter Server Due to Datastore Space Constraints — Broadcom Knowledge Base 405611, Updated On: 08-01-2025, accessed 20 November 2026
- vCert - Scripted vCenter expired certificate replacement — Broadcom Knowledge Base 385107, Updated On: 11-19-2026, accessed 20 November 2026
This article was originally published on .