Merge #137092 #137393

137092: kvnemesis: increase the retry backoff for aborted txns r=miraradeva a=miraradeva

In #135173, we tried introducing a backoff in the transaction retry mechanism (in `db.Txn`) in order to help with large transactions that repeatedly get in a deadlock (as seen in #133431). This wasn't enough as we continue to see this type of failure in #136266.

This commit increases the kvnemesis-specific retry backoff, which is meant to prevent thrashing of aborted transactions. We do see transactions getting aborted as part of deadlock resolution in both of the failures mentioned above, so hopefully this will help. Both failures above are rare and hard to repro.

Fixes: #136266

Release note: None

137393: roachprod: respect AWS_PROFILE env var r=golgeek,vidit-bhat a=herkolategan

Previously, if the AWS_PROFILE env var was set roachprod did not forward it to the `aws cli`. This change automatically sets the profile to the AWS_PROFILE env var if it is not specified via the `--aws-profile` roachprod flag.

Epic: None
Release note: None

Co-authored-by: Mira Radeva <[email protected]>
Co-authored-by: Herko Lategan <[email protected]>

pkg/kv/kvnemesis/applier.go

@@ -145,7 +145,7 @@ func applyOp(ctx context.Context, env *Env, db *kv.DB, op *Operation) {
 		// epochs of the same transaction to avoid waiting while holding locks.
 		retryOnAbort := retry.StartWithCtx(ctx, retry.Options{
 			InitialBackoff: 1 * time.Millisecond,
-			MaxBackoff:     250 * time.Millisecond,
+			MaxBackoff:     10 * time.Second,
 		})
 		var savedTxn *kv.Txn
 		txnErr := db.Txn(ctx, func(ctx context.Context, txn *kv.Txn) error {

pkg/roachprod/vm/aws/aws.go

@@ -496,7 +496,7 @@ func (o *ProviderOpts) ConfigureCreateFlags(flags *pflag.FlagSet) {
 
 // ConfigureClusterFlags implements vm.ProviderOpts.
 func (o *ProviderOpts) ConfigureClusterFlags(flags *pflag.FlagSet, _ vm.MultipleProjectsOption) {
-	flags.StringVar(&providerInstance.Profile, ProviderName+"-profile", providerInstance.Profile,
+	flags.StringVar(&providerInstance.Profile, ProviderName+"-profile", os.Getenv("AWS_PROFILE"),
 		"Profile to manage cluster in")
 	configFlagVal := awsConfigValue{awsConfig: *DefaultConfig}
 	providerInstance.Config = &configFlagVal.awsConfig