HDR failure handling on network disconnection
Posted in 2008
Topics: High Availability & Replication, Logging & Checkpoints, Networking & sqlhosts Configuration
As I know that HDR is working by HDR buffer for transferring primary data to secondary data. But what if the failure of network connection occurrs intermittently and the primary physical log was backed up to tape for freeing its space, how the informix data server handles it? Will the HDR buffer expend for keeping those outstanding data?
DANNY NG wrote:
> As I know that HDR is working by HDR buffer for transferring primary data to
> secondary data. But what if the failure of network connection occurrs
> intermittently and the primary physical log was backed up to tape for freeing
> its space, how the informix data server handles it? Will the HDR buffer
expend
> for keeping those outstanding data?
>
Danny,
First, you're a bit off. The physical log isn't ever backed up in the
way you imply. Second it has nothing to do with HDR. IB you are
thinking of the logical logs. In an HDR environment, if the secondary
is down, or the primary went down and the secondary took over, when the
down server comes back online, it tells the current primary what its
last logical log and log position was when it crashed and the primary
resends the logical log records that it missed if they are still
online. This happens even if the logical logs have been backed up
already and are marked free for reuse. When archived the logs are not
deleted, just marked as available to be overwritten. If the required
logs are no longer online, ie they have indeed been reused, then when
you restart the secondary it will report 'DR: Start Failure recovery
from tape'. If that happens then you have to roll forward the logs from
logical log archives at least to the point where the primary can pass
the remaining logical log records to the secondary (using ontape -l or
onbar -l) to complete the synchronization. If it's the primary thatfailed, the recovery is similar but may not be quite as automatic. See
the Administrator's Guide Chapter 20, the sections on 'Restarting If
Critical Data Is Not Damaged' which is on pages 20-26ff in the IDS 11.10
manuals (pages in earlier manual versions will differ).
Art S. Kagel
Is the process of logical log rolling forward by "ontape" or "onbar" are
automatic?