[Pgpool-general] pgpool: online recovery and serial field
Sergey V. Arlashin
s.arlashin at gmail.com
Sat Feb 5 16:42:42 UTC 2011
Hi!
I'm testing online recovery (multimaster replication) and there is an
issue related to it.
I made a simple script which writes data to my database during the whole
test. In the middle of the test I switch off one of the PostgreSQL nodes
and when it's back I start online recovery. Then I check the data on
both nodes. Everething is ok except fields which have 'serial' data
type. They are not the same. Here is an example:
node01:
id | data
------+-------
...
1293 | aaaaa
1294 | aaaaa
1295 | aaaaa
1296 | aaaaa
1297 | aaaaa
1298 | aaaaa
1299 | aaaaa
1300 | aaaaa
1301 | aaaaa
1333 | aaaaa
1334 | aaaaa
1335 | aaaaa
1336 | aaaaa
1337 | aaaaa
1338 | aaaaa
1339 | aaaaa
1340 | aaaaa
1341 | aaaaa
1342 | aaaaa
(1249 rows)
node02:
id | data
------+-------
...
1293 | aaaaa
1294 | aaaaa
1295 | aaaaa
1296 | aaaaa
1297 | aaaaa
1298 | aaaaa
1299 | aaaaa
1300 | aaaaa
1301 | aaaaa
1302 | aaaaa
1303 | aaaaa
1304 | aaaaa
1305 | aaaaa
1306 | aaaaa
1307 | aaaaa
1308 | aaaaa
1309 | aaaaa
1310 | aaaaa
1311 | aaaaa
(1249 rows)
As you see the number of rows on both nodes is the same, but values in
the id field (which has 'serial' data type) differ from each other. I
looked through pgpool documentation and found that 'insert_lock = true'
would probably solve my problem. But it didn't.
Could you tell me, is there a way to get fields with 'serial' data type
to be equal after online recovery on all nodes.
...
WBR, Sergey
More information about the Pgpool-general
mailing list