Thomas Tignor wrote: > We are experiencing intermittent DB corruption in postgres 9.5.14. We are > trying to > identify and eliminate all sources. We are using two independent services for > data > replication, Slony-I v2.2.6 and a custom service developed in-house. Both are > based > on COPY operations. DB corruption is observed when COPY operations fail with > an error > of the form: 'invalid byte sequence for encoding "UTF8"'. > This occurs with a frequency ranging between a few days and several weeks.
It might be a storage problem, but it could also be a problem with how the data get into the database in the first place. I would consider it unlikely that there is a bug in the TOAST code that changes just one byte. You could use block checksums to see if the storage is at fault, but unfortunately it requires an "initdb", dump and restore to switch to checksums. Yours, Laurenz Albe -- Cybertec | https://www.cybertec-postgresql.com
