There are no unadvertised utilities for maintaining replication entries. The replication feature in Accumulo has not been maintained since its inclusion, and there are many outstanding bugs such as these. It is because of these that we intend to mark it as deprecated in 2.1. I would strongly advise against relying on this feature, and consider alternatives to replicating via a pre-ingest pipeline rather than using Accumulo itself to handle your replication.
That said, what you've been doing to remove replication entries is probably a correct action, but you'll probably need to restart the cluster for the configuration changes to take effect so they don't keep getting recreated. If you have any log messages, especially with stack traces, we may be able to help further, but please note that there has been little interest in maintaining this feature to this point, and few people have expertise on this code. So, help may be limited. On Tue, Jul 12, 2022 at 3:42 PM Scott Kirklin <scott.kirk...@gmail.com> wrote: > > Hi, > > I am working with an accumulo instance that has over time collected quite a > few old and invalid entries in the replication table. Some rows correspond to > files that no longer exist, others are for tables that no longer exist, and > still others are for tables that no longer have replication targets > configured. I am curious if there are any unadvertised utilities or best > practices for maintaining the replication table when making schema or config > changes? > > For the final case, tables that no longer have replication targets defined, I > have tried just removing the replication table entries, but they continue to > be recreated even though the table definitely does not have any replication > targets defined (as evidenced by the continuous log warnings that No > configured targets for table with ID <id>). What am I missing to ensure those > are cleaned up permanently? > > Best, > Scott
