Update: VPS-RAM is staying
We are changing our earlier decision: VPS-RAM services will not be withdrawn. Instead, we are replacing the drives in xeon-1, the node they run on. The 15 December 2026 deadline described in The future of VPS-RAM services no longer applies, and there is no need to move your data to another service.
Drive replacement on xeon-1 (pl1)
Scheduled maintenance on the xeon-1 (pl1) node will begin on Sunday 15 November 2026 at around 02:00 CET. The NVMe drives will be replaced with enterprise SAS SSDs. This removes the cause of the earlier failures: NVMe drives that were not compatible with this platform.
- The maintenance is expected to take up to 6 hours.
- During this time VPS-RAM servers may be unavailable or restarted.
- All servers will be backed up before the work begins.
- Data on the servers will be preserved and IP addresses will stay the same.
Once the work is complete, VPS-RAM will be back on sale.
How it happened
On 7 November at 04:47 the xeon-1 node was shut down for scheduled maintenance covering work on its RAM and drives.
Before any hardware work we take a fresh backup of every service. After the node had been shut down, it turned out that the last backup job had not completed for all servers, which had been overlooked.
To finish the backup, we tried to start the node again. That was no longer possible: the platform reported an Uncorrectable PCI Express Error, and further diagnostics revealed failed controllers in both NVMe drives.
Why the failure was unexpected
Of the four drives in the node, the two that failed were both drives of a single RAID1 pool, in which each drive is a mirror copy of the other. The drives in that pool deliberately came from different series and were bought at different times, precisely to prevent this.
Attempts to recover data from the drives
To rule out a fault in the server itself, the drives were moved to another dedicated server built on a different platform (AMD Ryzen). They showed no signs of life there either.
We worked on two fronts at once. A diagnostic script kept trying to wake up the drive controllers, and in the meantime we bought new drives and started restoring data from backups onto new hardware. This way, bringing the services back did not depend on whether the data on the failed drives could be recovered.
Data recovery
The restore process included:
- 1
Reinstalling the node's operating system on new system drives.
- 2
Restoring services from two independent backup repositories.
- 3
Restoring services whose last backup had completed successfully to their state from the night of 6 to 7 November.
- 4
Restoring the remaining services from the backup of 6 November, taken about 24 hours earlier.
During the restore the server rebooted several times. Analysis of the hardware logs showed that the cause was other NVMe drives installed in the same expansion card. The data on those drives was safely copied to new drives, and the faulty drives were removed from the server together with the card. Since then the node has been running stably, and the remaining services were restored and started without interruption.
Data for all services has been recovered.
Compensation
We are aware that the outage, and in some cases the loss of changes from the last day, caused you real inconvenience.
As compensation, every customer has received 10 additional days on their billing period. The days have already been added to your services.
Lessons learned
The most important lesson from this incident: before any hardware work, including shutting a node down for scheduled maintenance, we must verify that the backup has completed successfully, without exception. Starting the job is not enough.
We have also removed the components that turned out to be the source of the instability from the node and protected its configuration against similar faults.
The future of VPS-RAM services
Updated on 10 November 2026
The decisions in this section have changed: VPS-RAM services will not be withdrawn and will be back on sale once the drives in the node have been replaced. See Update: VPS-RAM is staying.
All services are currently running stably. Recent events have shown, however, that with VPS-RAM we cannot guarantee the level of reliability you expect and that we ourselves want to provide. We do not want to risk a similar situation happening again, so we have made the following decisions:
Decisions about VPS-RAM services
- Sales of VPS-RAM services are suspended.
- Renewals of existing VPS-RAM services are suspended.
- VPS-RAM services will run until 15 December 2026.
VPS-RAM returns in Q2 2027
This pause is not permanent. VPS-RAM services will return to our offer in the second quarter of 2027, once we have found and tested stable solutions, above all for storage.
This decision applies to VPS-RAM services only. All other services continue unchanged and were not affected by the failure described above.
Please back up and move your data before that date. We suggest that VPS-RAM customers migrate to VPS-GAME, which runs stably with uptime measured in months. We will gladly help you move your data.
Migrate to VPS-GAME
We will help you move your data from VPS-RAM to the new service.
If your VPS-RAM service is paid for beyond 15 December 2026, please contact us by email at admin@pshost.pl or through the ticket system in the client panel. We will prepare an individual offer, such as a migration to VPS-GAME or a settlement for the unused period.
Thank you
Thank you for your patience, your understanding and all the messages of support we received during the outage. If you have any questions about your services, the recovered data or the next steps, please get in touch: