Ok, now create index is finished using maintenance_work_mem=100MB.
Thanks to all. I suppose that an error message more clear can help. Best Regards, Torello 2013/8/26 Torello Querci <tque...@gmail.com> > > > > 2013/8/26 Tom Lane <t...@sss.pgh.pa.us> > >> Torello Querci <tque...@gmail.com> writes: >> > 2013/8/26 Luca Ferrari <fluca1...@infinito.it> >> >> Is it possible to test with an incremented work_mem value? >> >> > Actually I use the default work_set value (1MB). >> >> maintenance_work_mem is what would be used for CREATE INDEX. >> >> Ok .... thanks > > >> FWIW, though, the combination of this weird error and the fact that you >> had a corrupt index to begin with makes me suspicious that there's some >> low-level problem. You might be well advised to do some memory testing >> on that machine, for example. >> >> I check for ecc memory but unfortunally the machine use non ecc memory. > This machine is installed on a remote site so I should to try to use e > memory tester in normal linux shell, so I can't use memtest at boot level. > > In this moment I get this error while executing the restore of the big > table in a different database on the same machine: > > psql:dump_ess_2013_08_26.sql:271177424: SSL error: sslv3 alert unexpected > message > psql:dump_ess_2013_08_26.sql:271177424: connection to server was lost > > I was connected with psql -h localhost. > > > Any other suggestion? > > Best Regards >
