Обсуждение: ERROR: subxact logged without previous toplevel record
Hi, I got the $SUBJECT error in a logical replication setup. Both versions are 10.3. I remembered that some bugs in the piece of code around this error were fixed in 9.6 but it seems not all of them. Unfortunately I don't have a test case. All DEBUG3 messages from an unsuccessful replication attempt are attached. The last few messages from the logical replication role are: 2018-05-15 15:24:05 -03 [31783]: [474-1] DEBUG: purged committed transactions from 63 to 63, xmin: 124517330, xmax: 124615171 2018-05-15 15:24:05 -03 [31783]: [475-1] DEBUG: xmin: 124517330, xmax: 124615171, oldestrunning: 124517330 2018-05-15 15:24:05 -03 [31783]: [476-1] DEBUG: failed to increase restart lsn: proposed 4C9/C9CC0938, after 4CA/8592A90, current candidate 4C9/C9CC0938, current after 4C9/D0FF05A8, flushed up to 4CA/9A8D138 2018-05-15 15:24:05 -03 [31783]: [477-1] DEBUG: spill 4096 changes in XID 124557507 to disk 2018-05-15 15:24:05 -03 [31783]: [478-1] ERROR: subxact logged without previous toplevel record -- Euler Taveira Timbira - http://www.timbira.com.br/ PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento
Вложения
On 2018-May-15, Euler Taveira wrote: > I got the $SUBJECT error in a logical replication setup. Both versions > are 10.3. I remembered that some bugs in the piece of code around this > error were fixed in 9.6 but it seems not all of them. Unfortunately I > don't have a test case. All DEBUG3 messages from an unsuccessful > replication attempt are attached. The last few messages from the > logical replication role are: > > 2018-05-15 15:24:05 -03 [31783]: [474-1] DEBUG: purged committed > transactions from 63 to 63, xmin: 124517330, xmax: 124615171 > 2018-05-15 15:24:05 -03 [31783]: [475-1] DEBUG: xmin: 124517330, > xmax: 124615171, oldestrunning: 124517330 > 2018-05-15 15:24:05 -03 [31783]: [476-1] DEBUG: failed to increase > restart lsn: proposed 4C9/C9CC0938, after 4CA/8592A90, current > candidate 4C9/C9CC0938, current after 4C9/D0FF05A8, flushed up to > 4CA/9A8D138 > 2018-05-15 15:24:05 -03 [31783]: [477-1] DEBUG: spill 4096 changes > in XID 124557507 to disk > 2018-05-15 15:24:05 -03 [31783]: [478-1] ERROR: subxact logged > without previous toplevel record Hello Can you find in pg_waldump all the records for this transaction, xid=124557507? -- Álvaro Herrera https://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
2018-05-15 16:08 GMT-03:00 Alvaro Herrera <alvherre@2ndquadrant.com>: > Can you find in pg_waldump all the records for this transaction, > xid=124557507? > I don't know this ML size limit but it seems huge to attach at this email (~900 kb gzipped). Grab it at [1]. [1] http://apinaye.timbira.com.br/~euler/tmp/txid-124557507.txt.gz -- Euler Taveira Timbira - http://www.timbira.com.br/ PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento
2018-05-15 16:43 GMT-03:00 Euler Taveira <euler@timbira.com.br>: > 2018-05-15 16:08 GMT-03:00 Alvaro Herrera <alvherre@2ndquadrant.com>: >> Can you find in pg_waldump all the records for this transaction, >> xid=124557507? >> > I don't know this ML size limit but it seems huge to attach at this > email (~900 kb gzipped). Grab it at [1]. > I forgot to mention that I have 16 replicas subscribed at the same publication. This problematic replica was setup only 10 minutes after the error. All of the other replicas are working fine since they were setup a few days ago. I am thinking if it is somewhat related to the replication start point. -- Euler Taveira Timbira - http://www.timbira.com.br/ PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento