-
Notifications
You must be signed in to change notification settings - Fork 222
Description
Hello,
Yesterday when testing with redis cluster, my application had been panic at line ( (of rueidis v1.0.46) :
return c.cs.s which in function Commands():
func (c *Completed) Commands() []string { return c.cs.s }
which is call by err = writeCmd(p.w, cmd.Commands())

The complete code is:
`func (p *pipe) _backgroundWrite() (err error) {
var (
ones = make([]Completed, 1)
multi []Completed
ch chan RedisResult
flushDelay = p.maxFlushDelay
flushStart = time.Time{}
)
for err == nil {
if ones[0], multi, ch = p.queue.NextWriteCmd(); ch == nil {
if flushDelay != 0 {
flushStart = time.Now()
}
if p.w.Buffered() != 0 {
if err = p.w.Flush(); err != nil {
break
}
}
ones[0], multi, ch = p.queue.WaitForWrite()
if flushDelay != 0 && atomic.LoadInt32(&p.waits) > 1 { // do not delay for sequential usage
// Blocking commands are executed in dedicated client which is acquired from pool.
// So, there is no sense to wait other commands to be written.
// https://github.com/redis/rueidis/issues/379
var blocked bool
for i := 0; i < len(multi) && !blocked; i++ {
blocked = multi[i].IsBlock()
}
if !blocked {
time.Sleep(flushDelay - time.Since(flushStart)) // ref: https://github.com/redis/rueidis/issues/156
}
}
}
if ch != nil && multi == nil {
multi = ones
}
for _, cmd := range multi {
err = writeCmd(p.w, cmd.Commands())
}
}
return
}`
Redis topology:
cluster enabled
3 master 3 replica
Action: Node is starting.
Could you help on this? Thanks. It seems cmd passed to writeCmd() has something wrong that lead to the crash. It confused because cmd is read from range multi, so I think something wrong happened to multi.