bdgranger commented on issue #9274: URL: https://github.com/apache/trafficserver/issues/9274#issuecomment-1363305623
As stated, this doesn't happen frequently. We often see instances where the rww reader is able to find the desired fragment right after the writer has written it: ``` [Nov 17 01:48:42.518] [ET_NET 39] DEBUG: (cache_read_agg) 0x2b6574065840: key: 48CB1F85 fKey: 2088A674 ReadDone doc key (50BA0878) doesn't match. Will probe again [Nov 17 01:48:42.519] [ET_NET 39] DEBUG: (cache_read_agg) 0x2b6574065840: key: 48CB1F85 fKey: 2088A674 ReadRead retrying(2): 190563 [Nov 17 01:48:42.606] [ET_NET 0] DEBUG: (cache_reenable) reenable 0x2b65b022a300 [Nov 17 01:48:42.606] [ET_NET 0] DEBUG: (cache_insert) 0x2b65b022a300: WriteDone: 48CB1F85, 2088A674, 10130 [Nov 17 01:48:42.616] [ET_NET 39] DEBUG: (cache_read_agg) 0x2b6574065840: key: 48CB1F85 fKey: 2088A674 ReadDone doc key (50BA0878) doesn't match. Will probe again [Nov 17 01:48:42.616] [ET_NET 39] DEBUG: (cache_read) 0x2b6574065840: Read complete on fragment 45C2571A851FCB488A2A6AD5B83B6464. Length: data payload=10130 this fragment=10202 total doc=200693 prefix=72 [Nov 17 01:48:42.616] [ET_NET 39] DEBUG: (cache_read_agg) 0x2b6574065840: key: 48CB1F85 fKey: 2088A674 ReadDone last_collision: 0x2b6c28de8e5c, dir_valid: 1 [Nov 17 01:48:42.616] [ET_NET 39] DEBUG: (cache_read_agg) 0x2b6574065840: key: C88890FE fKey: 2088A674 ReadMain retrying(0): 200693 [Nov 17 01:48:42.616] [ET_NET 39] DEBUG: (cache_reenable) reenable 0x2b6574065840 ``` So the question is why do we occasionally see this "stall" where the rww reader(s) cannot seem to find the fragment in the hash bucket linked list and just end up timing out? We have seen multiple rww readers all get stuck at the same fragment if they connect before that fragment was written, so it does not appear to be an issue specifically with any one reader, but an issue with dir_probe finding the next fragment in the linked list. Although eventually the fragment does become available. We have tried to enable dir_probe.* and dir_insert debug tags, but the problem does not appear to reproduce with that much logging enabled. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
