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

Fix Get EC object from non container node #1253

Merged
dstepanov-yadro merged 1 commit from dstepanov-yadro/frostfs-node:fix/ec_get_non_container_node into master 2024-09-04 19:51:10 +00:00
Conversation 0 Commits 1 Files changed 1 +6 -1
dstepanov-yadro commented 2024-07-17 06:09:27 +00:00
Member
Copy link

How it was: first, we collected information about the chunks of the EC object, taking into account the TTL, then we requested the chunks. This led to the fact that there was not enough TTL to collect information about the chunks: client -> node1 (non container, ttl=2) -> node2 (container, ttl=1), node2 can't get other chunks because of ttl=1.
How it became: first we get information that the object is an EU object, and then we collect information about the chunks and the chunks themselves in separate requests from the container node.

How it was: first, we collected information about the chunks of the EC object, taking into account the TTL, then we requested the chunks. This led to the fact that there was not enough TTL to collect information about the chunks: client -> node1 (non container, ttl=2) -> node2 (container, ttl=1), node2 can't get other chunks because of ttl=1. How it became: first we get information that the object is an EU object, and then we collect information about the chunks and the chunks themselves in separate requests from the container node.
dstepanov-yadro force-pushed fix/ec_get_non_container_node from 19db4bd93d to 795490b790 2024-07-17 06:11:28 +00:00 Compare
dstepanov-yadro force-pushed fix/ec_get_non_container_node from 795490b790 to ead691417b 2024-07-17 06:12:35 +00:00 Compare
dstepanov-yadro reviewed 2024-07-17 06:24:09 +00:00
pkg/services/object/delete/util.go Outdated
@ -32,2 +36,3 @@
}
p := getsvc.HeadPrm{}
p.SetCommonParameters(exec.commonParameters())
p.SetCommonParameters(&commonParameters)
dstepanov-yadro commented 2024-07-17 06:24:09 +00:00
Author
Member
Copy link

commonParameters may be changed during Head request (ForgetTokens, WithLocalOnly), so need a copy. Founded with autotests.

`commonParameters` may be changed during Head request (`ForgetTokens`, `WithLocalOnly`), so need a copy. Founded with autotests.
fyrchik commented 2024-07-17 07:27:17 +00:00
Owner
Copy link

Previously it was possible to pass nil, now we pass a pointer to the empty struct.
This is a change in behaviour, can anything bad happen?

Previously it was possible to pass `nil`, now we pass a pointer to the empty struct. This is a change in behaviour, can anything bad happen?
fyrchik commented 2024-07-17 07:27:30 +00:00
Owner
Copy link

Previously it was possible to pass nil, now we pass a pointer to the empty struct.
This is a change in behaviour, can anything bad happen?

Previously it was possible to pass `nil`, now we pass a pointer to the empty struct. This is a change in behaviour, can anything bad happen?
dstepanov-yadro commented 2024-07-17 07:32:03 +00:00
Author
Member
Copy link

Fixed

Fixed
dstepanov-yadro reviewed 2024-07-17 06:29:51 +00:00
pkg/services/object/get/assembleec.go Outdated
@ -36,2 +37,4 @@
r.log.Debug(logs.GetTryingToAssembleTheECObject)
// initialize epoch number
ok := r.initEpoch()
dstepanov-yadro commented 2024-07-17 06:29:51 +00:00
Author
Member
Copy link

Logging and status change - inside r.initEpoch() method.

Logging and status change - inside `r.initEpoch()` method.
dstepanov-yadro force-pushed fix/ec_get_non_container_node from ead691417b to 84d58795ce 2024-07-17 06:55:46 +00:00 Compare
dstepanov-yadro force-pushed fix/ec_get_non_container_node from 84d58795ce to 47de122454 2024-07-17 07:05:46 +00:00 Compare
dstepanov-yadro force-pushed fix/ec_get_non_container_node from 47de122454 to acc6c917b1 2024-07-17 07:22:36 +00:00 Compare
dstepanov-yadro force-pushed fix/ec_get_non_container_node from acc6c917b1 to 6917d9b285 2024-07-17 07:31:51 +00:00 Compare
fyrchik reviewed 2024-07-17 07:37:57 +00:00
pkg/services/object/get/assemblerec.go Outdated
@ -104,2 +142,2 @@
parts := a.retrieveParts(ctx, false)
c, err := a.getConstructor()
objID := a.addr.Object()
trav, cnr, err := a.traverserGenerator.GenerateTraverser(a.addr.Container(), &objID, a.epoch)
fyrchik commented 2024-07-17 07:36:09 +00:00
Owner
Copy link

This part seems to be common for all reconstruct*() helpers. If we move it one level above, will it make the code cleaner?

trav, cnr, err := a.traverserGenerator.GenerateTraverser(a.addr.Container(), &objID, a.epoch)
if err != nil {
	return nil, err
}
c, err := a.getConstructor(cnr)
if err != nil {
	return nil, err
}
This part seems to be common for all `reconstruct*()` helpers. If we move it one level above, will it make the code cleaner? ``` trav, cnr, err := a.traverserGenerator.GenerateTraverser(a.addr.Container(), &objID, a.epoch) if err != nil { return nil, err } c, err := a.getConstructor(cnr) if err != nil { return nil, err } ```
dstepanov-yadro commented 2024-07-17 08:04:04 +00:00
Author
Member
Copy link

Done

Done
pkg/services/object/get/assemblerec.go Outdated
@ -156,0 +200,4 @@
chunksGuard.Lock()
defer chunksGuard.Unlock()
for _, ch := range nodeChunks {
chunks[string(ch.ID.GetValue())] = ch
fyrchik commented 2024-07-17 07:37:06 +00:00
Owner
Copy link

We can override the old value here, does it matter?

We can override the old value here, does it matter?
fyrchik commented 2024-07-17 07:37:14 +00:00
Owner
Copy link

We can override the old value here, does it matter?

We can override the old value here, does it matter?
fyrchik commented 2024-07-17 07:37:22 +00:00
Owner
Copy link

We can override the old value here, does it matter?

We can override the old value here, does it matter?
fyrchik commented 2024-07-17 07:37:27 +00:00
Owner
Copy link

We can override the old value here, does it matter?

We can override the old value here, does it matter?
fyrchik commented 2024-07-17 07:37:27 +00:00
Owner
Copy link

We can override the old value here, does it matter?

We can override the old value here, does it matter?
fyrchik commented 2024-07-17 07:37:29 +00:00
Owner
Copy link

We can override the old value here, does it matter?

We can override the old value here, does it matter?
fyrchik commented 2024-07-17 07:37:30 +00:00
Owner
Copy link

We can override the old value here, does it matter?

We can override the old value here, does it matter?
fyrchik commented 2024-07-17 07:37:31 +00:00
Owner
Copy link

We can override the old value here, does it matter?

We can override the old value here, does it matter?
fyrchik commented 2024-07-17 07:37:48 +00:00
Owner
Copy link

We can override the old value here, does it matter?

We can override the old value here, does it matter?
dstepanov-yadro commented 2024-07-17 07:50:18 +00:00
Author
Member
Copy link

ch.ID is objectID. In certain situations, you can get the same chunk from different nodes.

`ch.ID` is objectID. In certain situations, you can get the same chunk from different nodes.
fyrchik approved these changes 2024-07-17 07:43:19 +00:00
pkg/services/object/get/assemblerec.go Outdated
@ -159,1 +233,4 @@
}
func (a *assemblerec) processECNodesRequests(ctx context.Context, nodes []placement.Node, dataCount, parityCount int) ([]*objectSDK.Object, error) {
foundedChunks := make(map[uint32]*objectSDK.Object)
fyrchik commented 2024-07-17 07:39:52 +00:00
Owner
Copy link

s/founded/found, founded is what happened with Saint-Petersburg in 1703 :)

`s/founded/found`, founded is what happened with Saint-Petersburg in 1703 :)
dstepanov-yadro commented 2024-07-17 08:03:16 +00:00
Author
Member
Copy link

Fixed

Fixed
pkg/services/object/get/get_test.go Outdated
@ -474,6 +478,7 @@ func testNodeMatrix(t testing.TB, dim []int) ([][]netmap.NodeInfo, [][]string) {
var ni netmap.NodeInfo
ni.SetNetworkEndpoints(a)
ni.SetPublicKey([]byte(a))
fyrchik commented 2024-07-17 07:42:22 +00:00
Owner
Copy link

Why is it important?

Why is it important?
fyrchik commented 2024-07-17 07:42:25 +00:00
Owner
Copy link

Why is it important?

Why is it important?
dstepanov-yadro commented 2024-07-17 07:56:50 +00:00
Author
Member
Copy link

Get service uses node's public key as map key for EC chunks. Without this change all chunks belong to the same node without public key in unit tests, so unit test fails.

`Get` service uses node's public key as map key for EC chunks. Without this change all chunks belong to the same node without public key in unit tests, so unit test fails.
dstepanov-yadro force-pushed fix/ec_get_non_container_node from 6917d9b285 to 0d76caf8df 2024-07-17 08:02:57 +00:00 Compare
aarifullin reviewed 2024-07-17 11:15:00 +00:00
pkg/services/object/get/assemblerec.go Outdated
@ -160,0 +257,4 @@
return ctx.Err()
default:
}
chunks := a.tryGetChunkListFromNode(ctx, info)
aarifullin commented 2024-07-17 11:15:00 +00:00
Member
Copy link

Just for my curiosity:

We get the list of chunks available on the observing node - could these sets of chunks be intersected by some reason?
It still works as you're using foundChunks map but can be some chunks is gotten by tryGetChunkFromRemoteStorage a few times?

Just for my curiosity: We get the list of chunks available on the observing `node` - could these sets of `chunks` be intersected by some reason? It still works as you're using `foundChunks` map but can be some chunks is gotten by `tryGetChunkFromRemoteStorage` a few times?
dstepanov-yadro commented 2024-07-17 11:17:57 +00:00
Author
Member
Copy link

Yes, in case of remove shard/return shard scenario. But it looks like rare case.

Yes, in case of remove shard/return shard scenario. But it looks like rare case.
👍 1
aarifullin approved these changes 2024-07-17 11:19:53 +00:00
dstepanov-yadro force-pushed fix/ec_get_non_container_node from 0d76caf8df to e83d39e33f 2024-07-17 11:25:00 +00:00 Compare
fyrchik approved these changes 2024-07-17 12:08:03 +00:00
pkg/services/object/get/assemblerec.go
@ -156,0 +203,4 @@
dataCount := policy.ECDataCount(cnr.Value.PlacementPolicy())
parityCount := policy.ECParityCount(cnr.Value.PlacementPolicy())
remoteNodes := make([]placement.Node, 0)
fyrchik commented 2024-07-17 12:06:18 +00:00
Owner
Copy link

var remoteNodes []placement.Node?

`var remoteNodes []placement.Node`?
dstepanov-yadro merged commit e83d39e33f into master 2024-07-17 14:58:31 +00:00
dstepanov-yadro deleted branch fix/ec_get_non_container_node 2024-07-17 14:59:15 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
aarifullin
fyrchik
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
3 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#1253
No description provided.
Delete branch "dstepanov-yadro/frostfs-node:fix/ec_get_non_container_node"

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?