ridv/gorealis
1
0
Fork 0
Code Issues Pull requests Projects Releases 23 Packages Wiki Activity Actions

Bugfix: switch statements were missing fallthrough statement thus making them retry non-retriable errors. Response is now propagated back up to caller if it's available in order to provide context if needed.

Browse source
This commit is contained in:
Renan DelValle 2018-10-03 16:51:22 -07:00
parent 9ebf118e71
commit 01911e47dd
No known key found for this signature in database
GPG key ID: C240AD6D6F443EC9
2 changed files with 10 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

4
realis.go
View file

@ -625,7 +625,7 @@ func (r *realisClient) CreateService(auroraJob Job, settings *aurora.JobUpdateSe
resp, err := r.StartJobUpdate(update, "")
if err != nil {
return nil, nil, errors.Wrap(err, "unable to create service")
return resp, nil, errors.Wrap(err, "unable to create service")
}
if resp != nil && resp.GetResult_() != nil {
@ -734,7 +734,7 @@ func (r *realisClient) StartJobUpdate(updateJob *UpdateJob, message string) (*au
})
if retryErr != nil {
return nil, errors.Wrap(retryErr, "Error sending StartJobUpdate command to Aurora Scheduler")
return resp, errors.Wrap(retryErr, "Error sending StartJobUpdate command to Aurora Scheduler")
}
return resp, nil
}

14
retry.go
View file

@ -15,9 +15,8 @@
package realis
import (
"time"
"math/rand"
"time"
"github.com/paypal/gorealis/gen-go/apache/aurora"
"github.com/paypal/gorealis/response"
@ -178,18 +177,21 @@ func (r *realisClient) thriftCallWithRetries(thriftCall auroraThriftCall) (*auro
case aurora.ResponseCode_OK:
return resp, nil
// If the response code is transient, continue retrying
// If the response code is transient, continue retrying
case aurora.ResponseCode_ERROR_TRANSIENT:
r.logger.Println("Aurora replied with Transient error code, retrying")
continue
// Failure scenarios, these indicate a bad payload or a bad config. Stop retrying.
// Failure scenarios, these indicate a bad payload or a bad config. Stop retrying.
case aurora.ResponseCode_INVALID_REQUEST:
fallthrough
case aurora.ResponseCode_ERROR:
fallthrough
case aurora.ResponseCode_AUTH_FAILED:
fallthrough
case aurora.ResponseCode_JOB_UPDATING_ERROR:
r.logger.Println("Terminal bad reply from Aurora, won't retry")
return nil, errors.New(response.CombineMessage(resp))
r.logger.Printf("Terminal Response Code %v from Aurora, won't retry\n", resp.GetResponseCode().String())
return resp, errors.New(response.CombineMessage(resp))
// The only case that should fall down to here is a WARNING response code.
// It is currently not used as a response in the scheduler so it is unknown how to handle it.