Closed yazun closed 3 years ago
There are few occurrences of this error log in the code, this is the datanode msg:
heap_prepare_freeze_tuple, heapam.c:7325 XX000: tuple cannot be frozen now, please try later xid 1165357 cutoff xid 1176649 committs 14511534806624 RecentDataTs 14496176693179 RecentGlobalXmin 1186649 RecentGlobalDataXmin 1186649#0122020-11-26 12:30:01 CET [10465,coord(0,0)]:xid[0-97/1178] [2-1] user=dr3_ops_cs36,db=surveys,client=192.168.168.155,query=psql
the later moment never arrives..
Further info
select min(xmin::text::bigint),max(xmax::text::bigint) from tbl; -- the tbl that cannot be frozen.
# min max
1 363986 1163851
select txid_current();
txid_current
1 867659
Update:
Lowering vacuum_delta to 10000 helps with vacuum freeze problem (as we see committs
was in the future in comparison with RecentDataTs
), does not help with the other:
ERROR: XX000 - node:datanode1, backend_pid:39522, nodename:datanode1,backend_pid:39522,message:relation 523409537 deleted while still in use
on create table .. as select...
Will start a new issue with the above problem.
We had to set higher (60000) vacuum_delta to avoid #39
Then, when we try to
vacuum freeze
we get on:Are these two related? It's a blocker as it also causes other bugs to appear, i.e. errors like
when trying to use a function in a
create tbl as ... fn_immutable_that_does_lookup(..) in qual
which are total blockers for us.Any hints?