Patch releases within a version are applied in the cache’s
maintenance window. Moving to a newer version, such as Valkey 8.1 to 9.1,
is your choice, made in one of two ways. Both are one-way: a cache can’t go back to an older version,
and a downgrade is refused with 409 downgrade_not_supported.
In place
ezgh caches update cch_4k2m9x0a7qpe --engine-version 9.1 --waitIn the console, it’s Upgrade on the cache’s Details tab. The cache is upgrading and keeps serving. It
starts at once, on caches with or without a standby. When it moves over, writes pause for a few
seconds and open connections end; clients reconnect to the same endpoint. No data is lost.
If the upgrade can’t complete, its operation fails and the cache stays on its version, unchanged.
Memcached: a new engineVersion is applied in the maintenance window, and the cache starts empty.
Blue/green deployments
Valkey only (409 not_supported on Memcached). A blue/green deployment makes a copy of the cache
(blue) on the new version (green), kept in step with blue, with its own endpoint to test
against. A switchover then makes green the cache, at blue’s endpoint and with its ID.
Create a deployment
ezgh caches blue-green create cch_4k2m9x0a7qpe --target-engine-version 9.1 --wait| Setting | |
|---|---|
targetEngineVersion |
The newer version green runs. |
targetInstanceType |
Green’s size. The cache’s by default. |
targetMode |
Green’s mode, memory or persistent. The cache’s by default. This is how a cache changes mode. |
targetStorage |
Green’s storage, for a persistent green. |
switchoverTimeoutSeconds |
How long a switchover may wait for green to catch up, 1 to 300 seconds. 30 by default. |
A cache has one open deployment at a time. Green gets the cache’s standby setting and its users.
While it’s open
- Green is read-only, on its own endpoint (
greenEndpoint,nulluntil the name resolves). Test your clients and reads against it. - Users you create, change, rotate or delete apply to blue and green.
- Other changes to the cache are refused with
409 blue_green_in_progress, except its name, description, tags and allowed sources, and maintenance waits. - A failover of blue keeps green in step with it.
Switch over
ezgh caches blue-green switchover cbg_6r2k8m4p1xqz --waitWrites pause on blue until green has every write, then the cache’s endpoint moves to green and green takes writes. Open connections end, and clients reconnect to the same endpoint. No data is lost. The cache keeps its ID, name and endpoint, and runs the new version. Green’s test endpoint goes away.
If green hasn’t caught up within the switchover timeout, blue takes writes again and nothing changes:
the switchover’s operation fails, and the deployment is available again.
A switchover is refused with 409 not_ready until the deployment is available. It’s one-way: the old
copy is gone once it completes.
Delete a deployment
ezgh caches blue-green delete cbg_6r2k8m4p1xqz --waitBefore a switchover, deleting discards green. After one, it removes the finished deployment.
Status
| Status | Means |
|---|---|
creating |
Green is being made. |
available |
Green is ready to test and to switch over to. |
switching_over |
A switchover is running. |
switched_over |
Green is the cache. |
failed |
The deployment ended; failure says why. Delete it. |
deleting |
It’s being deleted. |