TrueCloudLab
/
frostfs-node
18
1
Fork 20
Code Issues 88 Pull Requests 8 Actions Releases Activity

engine: Set Disabled mode to deleted shard #452

Merged
fyrchik merged 1 commits from dstepanov-yadro/frostfs-node:fix/read_after_close_support into support/v0.36 2023-06-20 07:43:07 +00:00
Conversation 0 Commits 1 Files Changed 10
dstepanov-yadro commented 2023-06-19 15:36:57 +00:00
Collaborator

To reproduce:

  1. Using frostfs-cli set shard mode to degraded-read-only.
  2. Set shard mode to disabled in config file, then SIGHUP.

The following may happen:

  • get request goroutine: gets shard list to search object
  • sighup goroutine: gets shard list to close, closes shards
  • get request goroutine: executes the request on the closed shard, blobovniza opens
To reproduce: 1. Using `frostfs-cli` set shard mode to `degraded-read-only`. 2. Set shard mode to `disabled` in config file, then SIGHUP. The following may happen: * get request goroutine: gets shard list to search object * sighup goroutine: gets shard list to close, closes shards * get request goroutine: executes the request on the closed shard, blobovniza opens
dstepanov-yadro force-pushed fix/read_after_close_support from 3898ec28a2 to 6ccb9b24a7 2023-06-19 15:37:56 +00:00 Compare
dstepanov-yadro requested review from storage-core-committers 2023-06-19 15:38:04 +00:00
dstepanov-yadro requested review from storage-core-developers 2023-06-19 15:38:04 +00:00
dstepanov-yadro requested review from fyrchik 2023-06-19 15:38:15 +00:00
dstepanov-yadro requested review from JuliaKovshova 2023-06-19 15:38:26 +00:00
dstepanov-yadro reviewed 2023-06-19 15:42:27 +00:00
pkg/local_object_storage/engine/control.go Outdated
@ -140,3 +140,2 @@
func (e *StorageEngine) close(releasePools bool) error {
e.mtx.RLock()
defer e.mtx.RUnlock()
e.mtx.Lock()
dstepanov-yadro commented 2023-06-19 15:42:27 +00:00
Poster
Collaborator

I believe it is mistake so i changed type of lock

I believe it is mistake so i changed type of lock
fyrchik commented 2023-06-19 15:45:02 +00:00
Owner

It could be mistake, but may be not: mtx protects e.shards which is read here.
Do you have any particular scenario in mind?

It could be mistake, but may be not: `mtx` protects `e.shards` which is read here. Do you have any particular scenario in mind?
dstepanov-yadro commented 2023-06-19 15:51:18 +00:00
Poster
Collaborator

Hm, right. Reverted.

Hm, right. Reverted.
fyrchik marked this conversation as resolved
dstepanov-yadro reviewed 2023-06-19 15:43:11 +00:00
pkg/local_object_storage/engine/shards.go Outdated
@ -175,3 +175,3 @@
for _, sh := range ss {
err := sh.Close()
err := sh.Close(true)
dstepanov-yadro commented 2023-06-19 15:43:11 +00:00
Poster
Collaborator

Disabled mode sets only when shards are deleted fron engine

Disabled mode sets only when shards are deleted fron engine
fyrchik reviewed 2023-06-19 15:43:44 +00:00
pkg/local_object_storage/shard/control.go Outdated
@ -261,0 +260,4 @@
func (s *Shard) Close(disable bool) error {
var lastErr error
if disable {
if err := s.SetMode(mode.Disabled); err != nil {
fyrchik commented 2023-06-19 15:43:21 +00:00
Owner

We have exactly 1 case, where disable == true, why not doing it on the caller-side when needed?

We have exactly 1 case, where `disable == true`, why not doing it on the caller-side when needed?
dstepanov-yadro commented 2023-06-19 15:49:21 +00:00
Poster
Collaborator

Right. Fixed.

Right. Fixed.
fyrchik marked this conversation as resolved
dstepanov-yadro reviewed 2023-06-19 15:44:11 +00:00
pkg/local_object_storage/shard/errors.go Outdated
@ -7,3 +8,4 @@
apistatus "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/client/status"
)
var ErrShardDisabled = logicerr.New("shard disabled")
dstepanov-yadro commented 2023-06-19 15:44:11 +00:00
Poster
Collaborator

Error type is logic to not to count this kind of errors with error counter

Error type is logic to not to count this kind of errors with error counter
👍 1
dstepanov-yadro force-pushed fix/read_after_close_support from 6ccb9b24a7 to 6867742c5c 2023-06-19 15:49:08 +00:00 Compare
dstepanov-yadro force-pushed fix/read_after_close_support from 6867742c5c to d015178f2d 2023-06-19 15:50:42 +00:00 Compare
dstepanov-yadro reviewed 2023-06-19 15:53:48 +00:00
pkg/local_object_storage/shard/tree.go Outdated
@ -160,2 +160,4 @@
return 0, ErrPiloramaDisabled
}
s.m.RLock()
dstepanov-yadro commented 2023-06-19 15:53:48 +00:00
Poster
Collaborator

I think it must be mode check too.

I think it must be mode check too.
👍 1
dstepanov-yadro force-pushed fix/read_after_close_support from d015178f2d to 1cae03c47c 2023-06-19 16:04:57 +00:00 Compare
fyrchik approved these changes 2023-06-19 16:08:07 +00:00
acid-ant approved these changes 2023-06-19 18:55:03 +00:00
JuliaKovshova approved these changes 2023-06-20 07:18:16 +00:00
JuliaKovshova approved these changes 2023-06-20 07:23:41 +00:00
fyrchik merged commit 1cae03c47c into support/v0.36 2023-06-20 07:43:07 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
fyrchik
acid-ant
JuliaKovshova
TrueCloudLab/storage-core-developers
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 Label
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
Assignees
Clear assignees
No Assignees
4 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#452
There is no content yet.
Delete Branch "dstepanov-yadro/frostfs-node:fix/read_after_close_support"

Deleting a branch is permanent. Although the deleted branch may exist for a short time before cleaning up, in most cases it CANNOT be undone. Continue?