`frostfs-node` doesn't stop while writecache flushing #1201

New issue

Closed

opened 2024-06-25 13:54:11 +00:00 by dstepanov-yadro · 0 comments

dstepanov-yadro commented

2024-06-25 13:54:11 +00:00

Member

Expected Behavior

frostfs-node must stop writecache flush and stop

Current Behavior

frostfs-node doesn't stop while writecache flushing big objects

Possible Solution

Invoke cancel() without Lock here:

func (c *cache) Close() error {
	// We cannot lock mutex for the whole operation duration
	// because it is taken by some background workers, so `wg.Wait()` is done without modeMtx.
	c.modeMtx.Lock()
	if c.cancel != nil {
		c.cancel()
		c.cancel = nil
	}
	c.mode = mode.DegradedReadOnly // prevent new operations from being processed
	c.modeMtx.Unlock()

Steps to Reproduce (for bugs)

Fill writecache with big objects (1MB for example)
Try to restart systemctl restart frostfs-storage

## Expected Behavior `frostfs-node` must stop writecache flush and stop ## Current Behavior `frostfs-node` doesn't stop while writecache flushing **big** objects ## Possible Solution Invoke `cancel()` without `Lock` here: ``` func (c *cache) Close() error { // We cannot lock mutex for the whole operation duration // because it is taken by some background workers, so `wg.Wait()` is done without modeMtx. c.modeMtx.Lock() if c.cancel != nil { c.cancel() c.cancel = nil } c.mode = mode.DegradedReadOnly // prevent new operations from being processed c.modeMtx.Unlock() ``` ## Steps to Reproduce (for bugs) 1. Fill writecache with big objects (1MB for example) 2. Try to restart `systemctl restart frostfs-storage`