Hello,
every time we release a copy with onbar , the Database server gets blocked one
to three minutes. Is this normal ?
Our version is 12.10.FC3WE and the command is onbar -b -w
Thanks
↪ replying to JOSE MARI RUIZ
Original post:
Hello,
every time we release a copy with onbar , the Database server gets blocked one
to three minutes. Is this normal ?
Our version is 12.10.FC3WE and the command is onbar -b -w
Thank
Response:
The answer is possibly. At the start of every backup, the server has to
perform a blocking checkpoint. This is different than what should be most
checkpoints happening on a system. Most checkpoints would be what's called
interval or non-blocking, meaning that the system is still able to perform
write activity while the buffer pool flushing is happening, so there is
normally only a very small block time. Durning a blocking checkpoint, the
server blocks write activity for the duration of the entire checkpoint, which
would include the buffer pool flushing. So if you h ave large buffer pools
with high values for LRU min/max (which control page flushing between
checkpoints) then yes, it may be normal for the server to have to block that
long to complete all the IO required for the checkpoint. You should be able to
look at the output of onstat -g ckp to see if the checkpoint that's happening
at the start of the backups being kicked off is taking a long time because of
"flush" time (that's the name of the column in the -g ckp output, and it is
the time it took the server to flush all the dirty buffers in the buffer
pool/s to disk. Anyway, I'd start looking there to get a better idea if this
is normal or not.
Jacques Renaut
HCL Informix Advanced Support