TrueCloudLab/frostfs-node
20
1
Fork 27
Code Issues 100 Pull requests 4 Projects Releases 1 Activity Actions

grpc: Increase message limits #470

Merged
fyrchik merged 1 commit from dstepanov-yadro/frostfs-node:fix/grpc_message_size into master 2023-06-30 13:52:05 +00:00
Conversation 0 Commits 1 Files changed 1 +3 -1
1 changed files with 3 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
cmd/frostfs-node/grpc.go
View file

@ -17,11 +17,13 @@ import (
"google.golang.org/grpc/credentials"
)
const maxRecvMsgSize = 256 << 20
func initGRPC(c *cfg) {
var successCount int
grpcconfig.IterateEndpoints(c.appCfg, func(sc *grpcconfig.Config) {
serverOpts := []grpc.ServerOption{
grpc.MaxSendMsgSize(maxMsgSize),
grpc.MaxRecvMsgSize(maxRecvMsgSize),
acid-ant commented 2023-06-26 18:36:21 +00:00
Outdated
Review
Copy link

Why not to move it in config file?

Why not to move it in config file?
dstepanov-yadro commented 2023-06-28 06:08:33 +00:00
Outdated
Review
Copy link

If we transfer this parameter to the configuration file, then we can get a situation where different nodes have different values set. This will lead to problems with requests.

If we transfer this parameter to the configuration file, then we can get a situation where different nodes have different values set. This will lead to problems with requests.
acid-ant commented 2023-06-28 06:53:34 +00:00
Outdated
Review
Copy link

We will face this error (if it will be) anyway because it is impossible to upgrade all nodes at once, rolling upgrade is a usual case for us.

We will face this error (if it will be) anyway because it is impossible to upgrade all nodes at once, rolling upgrade is a usual case for us.
dstepanov-yadro commented 2023-06-28 12:47:49 +00:00
Outdated
Review
Copy link

Well, you are right.
We can prevent invalid configuration, but we can't prevent rolling upgrade.

Well, you are right. We can prevent invalid configuration, but we can't prevent rolling upgrade.
fyrchik commented 2023-06-28 12:48:33 +00:00
Outdated
Review
Copy link

We have increased message size, so no problems should appear when talking old -> new node.
For new -> old we still use default chunkSize.

We have _increased_ message size, so no problems should appear when talking `old -> new` node. For `new -> old` we still use default `chunkSize`.
grpc.ChainUnaryInterceptor(
metrics.NewUnaryServerInterceptor(),
tracing.NewUnaryServerInterceptor(),