TrueCloudLab
/
frostfs-node
19
1
Fork 20
Code Issues 93 Pull Requests 8 Actions Releases Activity

node: Improve shutdown #930

Merged
fyrchik merged 2 commits from acid-ant/frostfs-node:bugfix/improve-shutdown into master 2024-01-31 08:30:35 +00:00
Conversation 2 Commits 2 Files Changed 6
acid-ant commented 2024-01-25 06:30:45 +00:00
Collaborator

Stop internal activity of the gc by context.
Release task pool of the policer when context cancelled.

Signed-off-by: Anton Nikiforov an.nikiforov@yadro.com

Stop internal activity of the `gc` by context. Release task pool of the `policer` when context cancelled. Signed-off-by: Anton Nikiforov <an.nikiforov@yadro.com>
acid-ant force-pushed bugfix/improve-shutdown from 4ecf408e19 to 748b173fc3 2024-01-25 06:32:42 +00:00 Compare
fyrchik requested review from storage-core-committers 2024-01-25 07:25:30 +00:00
fyrchik requested review from storage-core-developers 2024-01-25 07:25:30 +00:00
fyrchik reviewed 2024-01-25 07:41:57 +00:00
pkg/local_object_storage/engine/shards.go
@ -339,1 +341,4 @@
return
default:
}
sh.NotificationChannel() <- ev
fyrchik commented 2024-01-25 07:27:57 +00:00
Owner

Shouldn't it be select with both this and <-ctx.Done() branches?
A separate select for ctx.Done() is ok, because this way we guarantee early exit, though a single select is not much worse.

Shouldn't it be `select` with both this and `<-ctx.Done()` branches? A separate `select` for `ctx.Done()` is ok, because this way we guarantee early exit, though a single select is not much worse.
acid-ant commented 2024-01-25 07:53:39 +00:00
Poster
Collaborator

Right, the idea was to exit earlier. And it looks like once NotificationChannel is buffered, we can use one select for this too statements.

Right, the idea was to exit earlier. And it looks like once `NotificationChannel` is buffered, we can use one `select` for this too statements.
pkg/local_object_storage/shard/control.go Outdated
@ -151,0 +150,4 @@
// Currently it is used only for listening for the new epoch event.
// It is ok to keep in buffer one event, it helps us properly react on cancelling of the context.
// In other case we need to keep the order of component shutdown, like closing morph before engine.
eventChan: make(chan Event, 1),
fyrchik commented 2024-01-25 07:29:00 +00:00
Owner

Is it necessary if we implement select suggestion from another comment?

Is it necessary if we implement `select` suggestion from another comment?
acid-ant commented 2024-01-25 08:01:19 +00:00
Poster
Collaborator

Yes, because there might be situation when notification received and at the same time context cancelled. In scope of this PR, routine which reading from channel might end earlier than routine which write in it - this will lead to block on write in channel.

Yes, because there might be situation when notification received and at the same time context cancelled. In scope of this PR, routine which reading from channel might end earlier than routine which write in it - this will lead to block on write in channel.
fyrchik commented 2024-01-26 10:14:55 +00:00
Owner

Hm, how exactly? If we have select with 2 branches and 1 is available (ctx.Done()), there is no way we will block

Hm, how exactly? If we have select with 2 branches and 1 is available (`ctx.Done()`), there is no way we will block
acid-ant commented 2024-01-26 10:57:34 +00:00
Poster
Collaborator

No more working from home, messed in your comments. Right, once select used for writing in eventChannel buffering here is redundant.

No more working from home, messed in your comments. Right, once `select` used for writing in `eventChannel` buffering here is redundant.
dstepanov-yadro commented 2024-01-25 07:52:35 +00:00
Collaborator

Does SIGHUP work correctly?

Does `SIGHUP` work correctly?
acid-ant force-pushed bugfix/improve-shutdown from 748b173fc3 to f01961bdf9 2024-01-25 08:03:30 +00:00 Compare
acid-ant commented 2024-01-25 08:22:24 +00:00
Poster
Collaborator

Does SIGHUP work correctly?

There are no changes in implementation for reaction on SIGHUP - for storage engine we are reconfiguring only shards.

> Does `SIGHUP` work correctly? There are no changes in implementation for reaction on SIGHUP - for `storage engine` we are reconfiguring only shards.
dstepanov-yadro approved these changes 2024-01-26 09:50:48 +00:00
acid-ant force-pushed bugfix/improve-shutdown from f01961bdf9 to ba7fa7eca1 2024-01-26 10:01:30 +00:00 Compare
acid-ant force-pushed bugfix/improve-shutdown from ba7fa7eca1 to 766ee97e81 2024-01-26 10:54:53 +00:00 Compare
acid-ant force-pushed bugfix/improve-shutdown from 766ee97e81 to 5f88bf9c77 2024-01-29 08:58:43 +00:00 Compare
acid-ant force-pushed bugfix/improve-shutdown from 5f88bf9c77 to 1f2c34318f 2024-01-30 07:14:12 +00:00 Compare
fyrchik approved these changes 2024-01-31 08:30:28 +00:00
fyrchik merged commit e3573de6db into master 2024-01-31 08:30:35 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
dstepanov-yadro
fyrchik
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
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#930
There is no content yet.
Delete Branch "acid-ant/frostfs-node:bugfix/improve-shutdown"

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?