Moti Cohen

Hi @madolson, Since your tests didn’t encounter the issue it doesn’t mean we can ignore a real problem that reproduced to us and potentially can happen to others. IMHO, Let’s...

> @moticless can you also post the reproduction? Sorry for misleading. We didn't reproduce the issue. There were some indications and evaluation that we might reach this issue as part...

Hi @madolson, I managed to reproduce OOM by `flushall async` database with many small strings (DB1) and right after filling it up with many big strings (DB2). That is, `flushall...

I like your suggestion of generalizing this feature and adding also UNLINK. Maybe also lazy-evict?

I did all of my tests on my local machine. Generally speakaing, the time it takes to release largely influenced by the size of the items being released. Consider my...

I would like to advise you with about some design issue I encounter: Today there are 3 different types of events that triggers pauseClients(): - PAUSE_DURING_FAILOVER - PAUSE_BY_CLIENT_COMMAND - PAUSE_DURING_SHUTDOWN...

Already today we support two flavors of pause, CLIENT_PAUSE_WRITE and CLIENT_PAUSE_ALL. Example to API call: - pauseClients(... , CLIENT_PAUSE_WRITE); The selected flavor is rather hidden, and not reflected in the...

> In the first glance I think it's a over optimization case, because if user set maxmemory-eviction-tenacity to 100, the write command will be blocked until used_memory become lower than...

@oranagra @madolson , All along, I am not super happy with the idea of blocking write clients if there is at least one pending lazyfree job. We might reach paralyzing...

Hi @jonathon2nd, The logic of entering tilt mode is rather simple: ``` void sentinelCheckTiltCondition(void) { mstime_t now = mstime(); mstime_t delta = now - sentinel.previous_time; if (delta < 0 ||...