Bare Metal Solution maintenance
This page provides an overview of the process for Bare Metal Solution maintenance events that require server downtime.
To ensure that your Bare Metal Solution environment remains reliable, secure, and up-to-date, Google Cloud occasionally performs scheduled maintenance on the underlying Bare Metal Solution hardware and infrastructure. We perform many updates while your servers are running. However, some updates require downtime. In either case, we notify you about the scheduled maintenance and alert you if you need to take any actions to prepare for the maintenance.
We recommend that you allow the maintenance to be performed in the scheduled window for the following reasons:
- If you receive a maintenance notification for your server, it means that your server is running in a degraded state and must be attended as soon as possible to keep it running well and reduce the risk of unexpected failure and downtime.
- Regular maintenance helps keep your server functional over a long period of time.
Bare Metal Solution performs the following two types of maintenance:
Infrastructure maintenance—Includes the maintenance of the networking devices and storage components at the data center.
We notify you when an infrastructure maintenance takes place. Infrastructure maintenance notifications are only informational and do not need you to respond.
Server maintenance—Includes server maintenance activities, such as hardware replacement, patches, upgrades, and configuration changes.
For maintenance activities that do not impact your server (such as replacement of a hot-swappable and redundant power supply), we send you informational notifications to which you do not need to respond.
However, to perform maintenance that affects your server and requires downtime (such as DIMM replacement), we need your permission.
To collaborate with you on such a maintenance activity, we create a maintenance event. A maintenance event lets you view the details of a maintenance activity and take an appropriate action, such as approving or rescheduling the maintenance.
How maintenance works
Maintenance takes a lot of planning and coordination. It's a two-sided operation and requires you and us to work together to keep your servers functioning well.
To complete maintenance efficiently with minimal delay, we recommend that you read about and understand this process.
The following flowchart shows the steps involved in the Bare Metal Solution maintenance process:
The following sections describe each of the steps in detail.
Google creates a maintenance event
Ideally, we create a maintenance event two to three weeks before your server is due for maintenance. We use this event to coordinate and confirm with you to prevent or minimize disruption to your Bare Metal Solution environment. The event contains all the details about the maintenance activity and steps you need to take to prepare for the maintenance.
You can view your maintenance events in the Google Cloud console.
We send you reminders about your upcoming maintenance events seven days, three days, and one day before the maintenance event with preparatory steps for the event to be successful.
User reviews the maintenance event
We recommend that you read the details and impact of the event carefully. After you've reviewed the maintenance event, you can take the following actions:
- If you're okay with the proposed window, approve the maintenance event.
- If you're not okay with the proposed window, reschedule the maintenance event to propose a new window.
After we receive your response, we schedule your maintenance event.
We do not recommend canceling a maintenance event. But, if you still need to cancel a maintenance event due to business-critical reasons, see Cancel a maintenance event.
User prepares for the maintenance event
You must complete the preparation as explained in Prepare for a maintenance event before the maintenance event starts.
You must also complete any additional steps explained in your maintenance event.
Maintenance event takes place
We notify you when the maintenance event starts and when the maintenance event ends. These notifications are informational and do not require you to respond.
If the maintenance does not complete as planned, we reach out to you about the issues and next steps, including options to continue the maintenance event or revert the server to its previous state.
If you are unreachable or unavailable to respond in time, we roll back the changes and return the server to its previous state.