fix(device_remove): hot remove of spdk bdev should drop descriptor #1553
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
bors try |
tryBuild succeeded: |
dsharma-dc
force-pushed
the
dev_rm_child_close
branch
from
670fd33 to
1aa5adf
Compare
Signed-off-by: Diwakar Sharma <[email protected]>
dsharma-dc
force-pushed
the
dev_rm_child_close
branch
from
1aa5adf to
0bd4b3b
Compare
|
bors try |
|
bors try |
tryAlready running a review |
dsharma-dc
changed the title
| // will send us here and keeping the device/descriptor isn't of | ||
| // any use. | ||
| debug!("{self:?}: dropping block device"); | ||
| self.device = None; |
There was a problem hiding this comment.
hmm if there's a hot removal this way, does this mean we don't go through the pause dance?
If so then we need to cater for that as well..
There was a problem hiding this comment.
yeah we don't do that pause here because child retire isn't done. There are more nuances involved here than that make me think we are better maintain status quo with this code. I'm inclined to park this change for now.
- A way to simulate hot remove is doing
bdev destroyon pool disk rather thanpool destroy. In this case, both kind of child(nvmf and spdk bdev) behave similarly and are unplugged fine, without going via retire path and pause. - The
pool destroyorreplica destroyare to be rather considered as fault. nvmf child receivesAdminCommandCompletionFailurefor these, deems the child faulted, retires it, and then triggersDeviceRemovedevent. For spdk bdev, we directly getDeviceRemovedin these cases without child having gone through retire. - Practically, 1 is more likely than 2 in actual system.
tryBuild failed: |
|
not pursuing at the moment. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
If a bdev replica(non-nvmf) is destroyed, the device removal path doesn't close the device as in case of nvmf child. As a result, the child state isn't set as Destroyed, which results in bdev descriptor being kept opened and hence spdk internally stucks into a ebusy state for that bdev.