Add sync status #146
Add sync status #146
Conversation
Codecov Report
@@ Coverage Diff @@
## master #146 +/- ##
==========================================
+ Coverage 28.54% 29.14% +0.60%
==========================================
Files 61 61
Lines 3714 3785 +71
Branches 420 425 +5
==========================================
+ Hits 1060 1103 +43
- Misses 2640 2668 +28
Partials 14 14
|
dolphin/task_manager/tasks/task.py
Outdated
| def sync_task(resource_type): | ||
|
|
||
| @decorator.decorator | ||
| def _sync_task(f, *a, **k): |
There was a problem hiding this comment.
Suggest to set bits for all resources from the producer side. Producer set all bits at once, and consumer for a specific resource just reset the bit for it's own.
There was a problem hiding this comment.
Done, please check again
dolphin/common/constants.py
Outdated
| VOLUME = 2 | ||
| ACCESS_INFO = 3 | ||
| DISK = 4 | ||
| ALERT_SOURCE = 5 |
There was a problem hiding this comment.
ACCESS_INFO and ALERT_SOURCE are not part of task manager .How do we check if there is some activty happening for these resources like ( update access-info , or remove alert source ) ??
There was a problem hiding this comment.
Removed all of the rest except STORAGE_DEVICE, POOL and VOLUME.
dolphin/api/v1/storages.py
Outdated
| 0, | ||
| len(constants.ResourceType) - 1, | ||
| constants.SyncStatus.SYNCING) | ||
| db.storage_update(ctxt, storage['id'], storage) |
There was a problem hiding this comment.
Think about the scenario that two sync_all rquest arived at the same time, let's say request_A and request_B, handle steps as below:
- request_A get storage info first, and then request_B get storage info too,
- request_A check sync_status is not syncing so update it to be syncing,
- request_B check sync_status is not syncing so it will update the sync status again, which is not correct.
So read and update sync_status need lock.
|
@ThisIsClark , we also need to think about the dead lock scenario, say one consumer crashed and the coresponding bit will neber be reset. |
Raised a issue: #165 |
What this PR does / why we need it:
Add sync status to avoid two sync tasks for the same device are running at the same time.
For a storage device, add a int(named sync_status) to indicate the sync status, each bit of this int indicates a resource.
When a sync task for a resource is running for a device, set the special bit of sync_status of this device to 1, and when sync is done, set the special bit back to 0.
When received request to sync resource of a device, check if all the bit of sync_status is 0(means no sync task is running for this device), if not, this request would not be handled.
Which issue this PR fixes (optional, in
fixes #<issue number>(, fixes #<issue_number>, ...)format, will close that issue when PR gets merged): fixes #Special notes for your reviewer:
Release note:
Test case: