On Tue, Jun 7, 2011 at 10:03 PM, Eric Sigler wrote:
> We haven't watched the WAL continuously, but we have noticed that the
> WAL file grows slowly in size over time between application restarts
> (around every 2 weeks). Currently, the WAL file for one of our DBs is
> around 40MB, we've seen it
On Tue, Jun 7, 2011 at 9:02 PM, Eric Sigler wrote:
> Hello!
>
> Does anyone know of a reason why we might be seeing SQLite transaction
> rollbacks that take between 60 and 240 seconds? (One particularly odd
> occurrence was almost 20 minutes long!) This doesn't seem to happen
> often, but when
> We're issuing "PRAGMA
> wal_checkpoint" to the open DB handle.
If you want guaranteed finish of the checkpoint (and thus not growing
WAL-file) you need to issue "PRAGMA wal_checkpoint(RESTART)".
Pavel
On Tue, Jun 7, 2011 at 10:36 PM, Eric Sigler wrote:
> So, should the WAL file shrink back
So, should the WAL file shrink back to 0 then? We're issuing "PRAGMA
wal_checkpoint" to the open DB handle.
-Eric
On Tue, Jun 7, 2011 at 7:06 PM, Pavel Ivanov wrote:
>> (Actually, that was another general question we had, should that WAL
>> file ever shrink during use? Why would it grow to tha
> (Actually, that was another general question we had, should that WAL
> file ever shrink during use? Why would it grow to that size at all?)
It shrinks, when the full checkpoint is completed successfully. Until
then it grows.
Pavel
On Tue, Jun 7, 2011 at 10:03 PM, Eric Sigler wrote:
> We ha
We haven't watched the WAL continuously, but we have noticed that the
WAL file grows slowly in size over time between application restarts
(around every 2 weeks). Currently, the WAL file for one of our DBs is
around 40MB, we've seen it grow up to 130MB or so. I'll try to catch
the WAL size and se
> DB file in WAL mode, checkpointing done every 5 seconds by separate
> thread in program
Depending on the mode of checkpointing you use it can fail if there
are some other reading or writing transactions in progress. And at the
time you observe very long rollback actual checkpointing happens
beca
On 8 Jun 2011, at 2:02am, Eric Sigler wrote:
> Does anyone know of a reason why we might be seeing SQLite transaction
> rollbacks that take between 60 and 240 seconds?
My initial thought was a faulty hard disk: bad sectors or a duff controller.
Given that you're running inside a VM, it might a
Hello!
Does anyone know of a reason why we might be seeing SQLite transaction
rollbacks that take between 60 and 240 seconds? (One particularly odd
occurrence was almost 20 minutes long!) This doesn't seem to happen
often, but when it does it's painful. During the rollback, the disk
is definite
9 matches
Mail list logo
