Roman Gershman
Roman Gershman
This has been due to short `--replication_timeout` flag. The workaround is to increase the timeout to `200000` or higher.
@kostasrim why did you close? I still see `bool ServerState::AllowInlineScheduling()` checking for journaling
I do not know if SEPH can replace Dash in Dragonfly but there is one idea in SEPH that may be relevant here. I wanted to introduce an additional smaller...
The difference between two approaches is not significant though more thinking is required. 1. Buddy segment will require extension of PhysicalBid space so we could support iterators and support stable...
(2) could be tricky with stable traversal - we rely that a segment id of an item can not go down, only up during splits, and if the table shrinks...
I actually do not see how we can iterate over l2 segments separately. the only thing we can do is to iterate over them together, i.e. if a segment has...
I think (1) is simpler to implement so I will try to pursue it as POC.
@adiholden suggested a simpler solution (but less memory savings) - should provide ~7% better utilization. She suggested to allocate stash buckets separately. for example, today we have 56 regular buckets...
@chakaz I am pretty sure it was not, I saw it two week ago again.
succeeded to reproduce it. Got this stackl trace file: ``` #0 __pthread_kill_implementation (threadid=279096101982240, signo=signo@entry=6, no_tid=no_tid@entry=0) at ./nptl/pthread_kill.c:44 #1 0x0000fdd61f4c7690 in __pthread_kill_internal (signo=6, threadid=) at ./nptl/pthread_kill.c:78 #2 0x0000fdd61f47cb3c in __GI_raise (sig=6)...