TrueCloudLab/frostfs-node
19
1
Fork 27
Code Issues 92 Pull requests 9 Projects Releases 1 Activity Actions

WIP: Process EC container for Get/GetRange/Head concurrently #1237

Closed
dstepanov-yadro wants to merge 6 commits from dstepanov-yadro/frostfs-node:fix/ec_get_failover into master
pull from: dstepanov-yadro/frostfs-node:fix/ec_get_failover
merge into: TrueCloudLab:master
Conversation 1 Commits 6 Files changed 15 +424 -117
dstepanov-yadro commented 2024-07-10 07:35:06 +00:00
Member
Copy link

Now get requests for EC containers run concurrently with concurrency limit equal to EC data count.

Now `get` requests for EC containers run concurrently with concurrency limit equal to EC data count.
dstepanov-yadro force-pushed fix/ec_get_failover from 32afe410d3 to c9b1af0635 2024-07-10 07:38:30 +00:00 Compare
dstepanov-yadro changed title from WIP: Process EC container for Get/GetRange/Head concurrently to Process EC container for Get/GetRange/Head concurrently 2024-07-10 09:07:29 +00:00
dstepanov-yadro force-pushed fix/ec_get_failover from c9b1af0635 to 30c51fbd5d 2024-07-10 09:07:32 +00:00 Compare
dstepanov-yadro requested review from storage-core-committers 2024-07-10 09:07:40 +00:00
dstepanov-yadro requested review from storage-core-developers 2024-07-10 09:07:47 +00:00
dstepanov-yadro force-pushed fix/ec_get_failover from 31e94be0b6 to e33752e034 2024-07-10 09:18:00 +00:00 Compare
acid-ant reviewed 2024-07-10 09:28:44 +00:00
acid-ant reviewed 2024-07-10 09:34:58 +00:00
pkg/services/object/get/container.go Outdated
@ -94,0 +149,4 @@
}()
err := r.processECNodesRequests(ctx, nodes, dataCount)
cancel()
acid-ant commented 2024-07-10 09:34:57 +00:00
Member
Copy link

You are canceling context here and using it below for r.writeCollectedObject(ctx), looks wrong.

You are canceling context here and using it below for `r.writeCollectedObject(ctx)`, looks wrong.
dstepanov-yadro commented 2024-07-10 09:40:18 +00:00
Author
Member
Copy link

Right, fixed.

Right, fixed.
acid-ant marked this conversation as resolved
dstepanov-yadro force-pushed fix/ec_get_failover from e33752e034 to c4d310d669 2024-07-10 09:38:27 +00:00 Compare
aarifullin reviewed 2024-07-10 10:33:19 +00:00
pkg/services/object/get/container.go Outdated
@ -94,0 +170,4 @@
defer cancel()
var wg sync.WaitGroup
wg.Add(1)
go func() {
aarifullin commented 2024-07-10 10:33:19 +00:00
Member
Copy link

[Optional]

What do you think: won't this approach cause goroutine spam if too many get-like requests are processed at the same time? Using errgroup would be safe I guess

[Optional] What do you think: won't this approach cause goroutine spam if too many get-like requests are processed at the same time? Using `errgroup` would be safe I guess
dstepanov-yadro commented 2024-07-10 12:28:23 +00:00
Author
Member
Copy link

I didn't catch the thought.
errgroup creates the same goroutines.

I didn't catch the thought. `errgroup` creates the same goroutines.
aarifullin commented 2024-07-10 12:54:15 +00:00
Member
Copy link

errgroup creates the same goroutines.

Yes, it does but with the fixed number of workers.

Nevermind, sorry. I incorrecly read this part:

go func() {
		defer wg.Done()

		for {
			batch := traverser.Next()
			if len(batch) == 0 {
				r.log.Debug(logs.NoMoreNodesAbortPlacementIteration)
				close(nodes)
				return
			}
			for _, node := range batch {
				select {
				case <-ctx.Done():
					return
				case nodes <- node:
				}
			}
		}
	}()

I've mistaken this goroutine generation within for-loop

> errgroup creates the same goroutines. Yes, it does but with the fixed number of workers. Nevermind, sorry. I incorrecly read this part: ```go go func() { defer wg.Done() for { batch := traverser.Next() if len(batch) == 0 { r.log.Debug(logs.NoMoreNodesAbortPlacementIteration) close(nodes) return } for _, node := range batch { select { case <-ctx.Done(): return case nodes <- node: } } } }() ``` I've mistaken this goroutine generation within for-loop
aarifullin marked this conversation as resolved
fyrchik reviewed 2024-07-10 11:22:24 +00:00
pkg/services/object/get/container.go Outdated
@ -55,1 +65,4 @@
// In most cases, the value from the cache will be returned.
// Container appears in the cache when traverser generates.
cnr, err := r.containerSource.Get(r.address().Container())
fyrchik commented 2024-07-10 11:06:51 +00:00
Owner
Copy link

Can we get it from the traverser directly?

Can we get it from the traverser directly?
dstepanov-yadro commented 2024-07-10 12:15:44 +00:00
Author
Member
Copy link

Container is not traverser part, but container is requered to generate traverser.
Fixed: now traverser generator returns container.

Container is not traverser part, but container is requered to generate traverser. Fixed: now traverser generator returns container.
pkg/services/object/get/container.go Outdated
@ -94,0 +129,4 @@
var errECInfo *objectSDK.ECInfoError
var errRemoved *apistatus.ObjectAlreadyRemoved
var errOutOfRange *apistatus.ObjectOutOfRange
var errSuccess *ecGetSuccessErr
fyrchik commented 2024-07-10 11:09:44 +00:00
Owner
Copy link

What is the benefit of using a separate error type for successfully performed requests?
Go has multiple value returns can we use them?

What is the benefit of using a separate error type for successfully performed requests? Go has multiple value returns can we use them?
dstepanov-yadro commented 2024-07-10 11:43:25 +00:00
Author
Member
Copy link

To cancel errgroup's context and do not run other requests: if eg.Go's argument returns error, then errgroup cancels context. Of course it is possible to make the same thing without errgroup, but it will be not so simple.
Also added ctx.Done check.

To cancel `errgroup`'s context and do not run other requests: if `eg.Go`'s argument returns error, then errgroup cancels context. Of course it is possible to make the same thing without errgroup, but it will be not so simple. Also added `ctx.Done` check.
fyrchik commented 2024-07-10 11:57:41 +00:00
Owner
Copy link

It can be done with a special error, e.g. errStopIteration is what we use somewhere.

It can be done with a special error, e.g. `errStopIteration` is what we use somewhere.
dstepanov-yadro commented 2024-07-10 12:01:32 +00:00
Author
Member
Copy link

But ecGetSuccessErr is special error used only to stop EC handling and pass result object.

But `ecGetSuccessErr` is special error used only to stop EC handling and pass result object.
pkg/services/object/get/container.go Outdated
@ -94,0 +132,4 @@
var errSuccess *ecGetSuccessErr
switch {
case err == nil: // nil returns if all nodes failed or incomplete EC info received
fyrchik commented 2024-07-10 11:08:23 +00:00
Owner
Copy link

s/returns/is returned/

`s/returns/is returned/`
dstepanov-yadro commented 2024-07-10 11:44:53 +00:00
Author
Member
Copy link

done

done
fyrchik marked this conversation as resolved
pkg/services/object/get/container.go Outdated
@ -94,0 +209,4 @@
r.log.Debug(logs.GetCouldNotConstructRemoteNodeClient, zap.String("node_key", hex.EncodeToString(info.PublicKey())))
return err
}
obj, err := r.getRemote(egCtx, rs, info)
fyrchik commented 2024-07-10 11:22:17 +00:00
Owner
Copy link

So for each node we call getRemote with the same parameters.
Where is the place in code where we do combine chunks in the final object?

So for each node we call `getRemote` with the same parameters. Where is the place in code where we do combine chunks in the final object?
dstepanov-yadro commented 2024-07-10 11:49:30 +00:00
Author
Member
Copy link

Line 122 in 3bf6e6d
func (a *assemblerec) retrieveParts(mainCtx context.Context, headOnly bool) []*objectSDK.Object {

https://git.frostfs.info/TrueCloudLab/frostfs-node/src/commit/3bf6e6dde60f15f4f90caf44eda429e8f6269cc1/pkg/services/object/get/assemblerec.go#L122
fyrchik commented 2024-07-10 12:08:34 +00:00
Owner
Copy link

I have realised where my confusion comes from: we have 2 steps:

  1. Get ECInfo from enough nodes
  2. Pull EC chunks

The "enough nodes" is defined on step 1, even though nodes may go down at step 2. In this case we will fail to fetch object?
The situation will happen during failover tests.

I have realised where my confusion comes from: we have 2 steps: 1. Get ECInfo from enough nodes 2. Pull EC chunks The "enough nodes" is defined on step 1, even though nodes may go down at step 2. In this case we will fail to fetch object? The situation _will_ happen during failover tests.
dstepanov-yadro commented 2024-07-10 12:13:15 +00:00
Author
Member
Copy link

Right, if node fails between step 1 and step 2, then such request will fail.

Right, if node fails between step 1 and step 2, then such request will fail.
pkg/services/object/get/container.go Outdated
@ -94,0 +230,4 @@
ecInfoGuard.Lock()
defer ecInfoGuard.Unlock()
r.infoEC = util.MergeECInfo(errECInfo.ECInfo(), r.infoEC)
if r.isRaw() {
fyrchik commented 2024-07-10 11:12:22 +00:00
Owner
Copy link

Could you explain this line a bit? Why do we NOT return error if the number of chunks is not equal to the expected?

Could you explain this line a bit? Why do we NOT return error if the number of chunks is not equal to the expected?
dstepanov-yadro commented 2024-07-10 11:46:23 +00:00
Author
Member
Copy link

In case of HEAD request with raw flag it is expected to get raw EC info error with all chunks.

In case of `HEAD` request with `raw` flag it is expected to get raw EC info error with all chunks.
dstepanov-yadro force-pushed fix/ec_get_failover from c4d310d669 to d731be8011 2024-07-10 11:39:49 +00:00 Compare
dstepanov-yadro force-pushed fix/ec_get_failover from d731be8011 to 7f7d04882a 2024-07-10 11:44:56 +00:00 Compare
elebedeva reviewed 2024-07-10 12:02:01 +00:00
pkg/services/object/get/container.go Outdated
@ -54,2 +64,4 @@
}
// In most cases, the value from the cache will be returned.
// Container appears in the cache when traverser generates.
elebedeva commented 2024-07-10 12:02:01 +00:00
Member
Copy link

Looks like a grammar mistake: traverser generates --> traverser is generated.

Looks like a grammar mistake: `traverser generates` --> `traverser is generated`.
👍 1
dstepanov-yadro commented 2024-07-10 12:13:37 +00:00
Author
Member
Copy link

Fixed

Fixed
elebedeva marked this conversation as resolved
dstepanov-yadro force-pushed fix/ec_get_failover from 99790c09a0 to e89f8bb6c7 2024-07-10 12:17:28 +00:00 Compare
dstepanov-yadro force-pushed fix/ec_get_failover from e89f8bb6c7 to eebf7b4bfb 2024-07-10 12:21:09 +00:00 Compare
aarifullin approved these changes 2024-07-10 12:55:30 +00:00
acid-ant reviewed 2024-07-10 13:21:39 +00:00
pkg/services/object/get/remote.go Outdated
@ -65,19 +65,10 @@ func (r *request) processNode(ctx context.Context, info client.NodeInfo) bool {
mergeSplitInfo(r.splitInfo(), errSplitInfo.SplitInfo())
r.err = objectSDK.NewSplitInfoError(r.infoSplit)
case errors.As(err, &errECInfo):
acid-ant commented 2024-07-10 13:21:39 +00:00
Member
Copy link

Looks like part of the code in the default section is no more needed:

		if r.status == statusEC {
			// we need to continue getting another chunks  from another nodes
...
Looks like part of the code in the `default` section is no more needed: ``` if r.status == statusEC { // we need to continue getting another chunks from another nodes ... ```
dstepanov-yadro commented 2024-07-10 13:25:33 +00:00
Author
Member
Copy link

done

done
acid-ant marked this conversation as resolved
dstepanov-yadro force-pushed fix/ec_get_failover from eebf7b4bfb to 5d36e847b0 2024-07-10 13:24:59 +00:00 Compare
acid-ant approved these changes 2024-07-10 14:15:36 +00:00
aarifullin approved these changes 2024-07-12 08:03:51 +00:00
dstepanov-yadro force-pushed fix/ec_get_failover from 5d36e847b0 to 906570789c 2024-07-12 08:44:28 +00:00 Compare
dstepanov-yadro force-pushed fix/ec_get_failover from 906570789c to 50b9dea0fb 2024-07-12 09:40:10 +00:00 Compare
achuprov approved these changes 2024-07-12 09:47:38 +00:00
aarifullin approved these changes 2024-07-12 12:52:35 +00:00
dstepanov-yadro changed title from Process EC container for Get/GetRange/Head concurrently to WIP: Process EC container for Get/GetRange/Head concurrently 2024-07-12 16:41:54 +00:00
dstepanov-yadro commented 2024-07-15 05:22:25 +00:00
Author
Member
Copy link

Incorrect implementation, close

Incorrect implementation, close
dstepanov-yadro closed this pull request 2024-07-15 05:22:40 +00:00
All checks were successful
Vulncheck / Vulncheck (pull_request) Successful in 4m16s
Required
Details
DCO action / DCO (pull_request) Successful in 4m3s
Required
Details
Build / Build Components (1.21) (pull_request) Successful in 5m25s
Required
Details
Build / Build Components (1.22) (pull_request) Successful in 5m33s
Required
Details
Tests and linters / gopls check (pull_request) Successful in 6m40s
Required
Details
Tests and linters / Staticcheck (pull_request) Successful in 7m32s
Required
Details
Tests and linters / Lint (pull_request) Successful in 8m20s
Required
Details
Pre-commit hooks / Pre-commit (pull_request) Successful in 11m15s
Required
Details
Tests and linters / Tests with -race (pull_request) Successful in 11m35s
Required
Details
Tests and linters / Tests (1.21) (pull_request) Successful in 11m50s
Required
Details
Tests and linters / Tests (1.22) (pull_request) Successful in 12m10s
Required
Details

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
Reviewers
No reviewers
acid-ant
achuprov
aarifullin
Labels
Clear labels   
P0

Highest

  
P1

High

  
P2

Medium

  
P3

Low

  
badger

Badger DB related issues

  
frostfs-adm

   
frostfs-cli

   
frostfs-ir

   
frostfs-lens

   
frostfs-node

   
good first issue

Good for newcomers

  
triage
  
Infrastructure

Infrastructure, Automation, CI/CD or DevOps related stuff

  
blocked

The issue or PR is blocked by something

  
bug

Something is not working

  
config

Configuration format update or breaking change

  
discussion

Talking about something in order to reach a decision or to exchange ideas

  
documentation

Documentation related

  
duplicate

This issue or pull request already exists

  
enhancement

New feature

  
go

Language features adoption

  
help wanted

Extra attention is needed

  
internal

   
invalid

Something is wrong

  
kludge

This is a kludge and we know it

  
observability

Related to metrics, tracing and logs

  
perfomance

Perfomance related

  
question

More information is needed

  
refactoring

Refactoring

  
wontfix

This won't be fixed

No labels
P0
P1
P2
P3
badger
frostfs-adm
frostfs-cli
frostfs-ir
frostfs-lens
frostfs-node
good first issue
triage
Infrastructure
blocked
bug
config
discussion
documentation
duplicate
enhancement
go
help wanted
internal
invalid
kludge
observability
perfomance
question
refactoring
wontfix
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
6 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: TrueCloudLab/frostfs-node#1237
No description provided.
Delete branch "dstepanov-yadro/frostfs-node:fix/ec_get_failover"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?