add retry backoff interval #3461
Conversation
Signed-off-by: Rama Chavali <[email protected]>
ramaraochavali
added
the
release-notes-none
istio-testing
added
the
size/XS
| @@ -1317,6 +1317,9 @@ message HTTPRetry { | |||
| // Flag to specify whether the retries should ignore previously tried hosts during retry. | |||
| // Defaults to true. | |||
| google.protobuf.BoolValue retry_ignore_previous_hosts = 5; | |||
|
|
|||
| // Specifies the minimum duration between retry attempts. | |||
There was a problem hiding this comment.
Can we indicate something along the lines of "if unset, retries will be instant"? And maybe clarify the interaction with a timeout (e.g. what if my backoff=2s, timeout=3s, retries=3 - what happens? I assume we can retry at most 1x?)
There was a problem hiding this comment.
Ok. Updated. PTAL
Signed-off-by: Rama Chavali <[email protected]>
|
@howardjohn can PTAL when you get chance? |
| @@ -1317,6 +1317,13 @@ message HTTPRetry { | |||
| // Flag to specify whether the retries should ignore previously tried hosts during retry. | |||
| // Defaults to true. | |||
| google.protobuf.BoolValue retry_ignore_previous_hosts = 5; | |||
|
|
|||
| // Specifies the minimum duration between retry attempts. | |||
| // If unset, default minimum duration of 25ms is used as base interval for exponetial backoff. | |||
There was a problem hiding this comment.
Isn't the default no backoff?
There was a problem hiding this comment.
There was a problem hiding this comment.
Huh. TIL. I always thought it was instant
There was a problem hiding this comment.
Is it valid to set 0s? I think our standard valdiation will reject it unless we add protoc-gen-crd:duration-validation:none.
Note the GW API says "if unset its implementation specific" so we can keep that at 25ms, but they do allow 0s i think
There was a problem hiding this comment.
it should be greater than 0s in Envoy. If gateway allows that, we will have to tweak that in Envoy.
Configures retry backoff interval. Needed for gateway api also