Hmm, did you disable and then remove the table quickly? My guess is that those 2 actions took place too quickly (we need to add a better way to block). The other way that could happen is if you have 2 tables pointed at the same directory.
On Wed, Apr 10, 2013 at 7:11 AM, Tim Williams <[email protected]> wrote: > After deleting a table that's been around a while, in the shard server > logs, I see tons of these errors: > > "IO Error during commit of [$tablename/shard-0000000N]"... "... no > longer has write lock." > > Anyone see this? Have an idea of how to tell the shard the table is > gone without bouncing the cluster? The table was created before I > realized I should be flipping the readOnly bit on table creation, if > that matters... > > Thanks, > --tim >
