Difference in checkpoint durations on HDR Primary and Secondary.

Hiya guys! I've just come across something I didn't think was possible: IDS 10.00.FC5 on Linux RedHat. Primary: ... 14:26:53 Checkpoint Completed: duration was 0 seconds. 14:26:53 Checkpoint loguniq 8922, logpos 0x2066018, timestamp:0x93d695b1 14:26:53 Maximum server connections 476 Secondary: ... 14:26:59 Checkpoint Completed: duration was 5 seconds. 14:26:59 Checkpoint loguniq 8922, logpos 0x2066018, timestamp:0xa439eb15 14:26:59 Maximum server connections 1 The time difference between the system clocks is on both boxes is unnoticeable (less than a second). It seems as though the checkpoint on the Primary isn't waiting for the checkpoint to complete on the HDR Secondary before completing, which I thought it should: " ... Checkpoints Between Database Servers Checkpoints between database servers in a replication pair are synchronous, regardless of the value of DRINTERVAL. (See "Checkpoints" on page 16-7.) A checkpoint on the primary database server completes only after it completes on the secondary database server. ... " (From "Version 10, IBM Informix Dynamic Server Administrator's Guide, G251-2267-02" , page 20-10) Could this be a new feature to keep checkpoints short on an HDR Primary? Then I'd guess this would be a documentation bug. Any thoughts? RoB

RoB wrote: > Hiya guys! > > I've just come across something I didn't think was possible: > > IDS 10.00.FC5 on Linux RedHat. > > > Primary: > > ... > 14:26:53 Checkpoint Completed: duration was 0 seconds. > 14:26:53 Checkpoint loguniq 8922, logpos 0x2066018, timestamp:> 0x93d695b1 > > 14:26:53 Maximum server connections 476> > > Secondary: > > ... > 14:26:59 Checkpoint Completed: duration was 5 seconds. > 14:26:59 Checkpoint loguniq 8922, logpos 0x2066018, timestamp:> 0xa439eb15 > > 14:26:59 Maximum server connections 1> > > > The time difference between the system clocks is on both boxes is > unnoticeable (less than a second). > > It seems as though the checkpoint on the Primary isn't waiting for > the checkpoint to complete on the HDR Secondary before completing, > which I thought it should: There's a bit of confusion about what we mean when we say "SYNC mode" in context with HDR. Basically what that means is that the flushing of the log buffer to disk is not considered complete (and available for reuse) until the log buffer has been shipped to the secondary server AND the shipment of the transmission has been ACKed. It doesn't mean that the log records have been processed on the secondary. > > " > ... > Checkpoints Between Database Servers > > Checkpoints between database servers in a replication pair are > synchronous, regardless of the value of DRINTERVAL. (See > "Checkpoints" on page 16-7.) A checkpoint on the primary database > server completes only after it completes on the secondary database > server. > ... > " > > (From "Version 10, IBM Informix Dynamic Server Administrator's > Guide, G251-2267-02" , page 20-10) > > Could this be a new feature to keep checkpoints short on an HDR > Primary? Then I'd guess this would be a documentation bug. > > Any thoughts? > > RoB >

Yes, that makes more sense. Although, the documentation isn't as clear on the issue as one would expect (it's actually a bit contradictive I would say). Thanks, RoB Madison Pruet wrote: > RoB wrote: > > Hiya guys! > > > > I've just come across something I didn't think was possible: > > > > IDS 10.00.FC5 on Linux RedHat. > > > > > > Primary: > > > > ... > > 14:26:53 Checkpoint Completed: duration was 0 seconds. > > 14:26:53 Checkpoint loguniq 8922, logpos 0x2066018, timestamp:> > 0x93d695b1 > > > > 14:26:53 Maximum server connections 476> > > > > > Secondary: > > > > ... > > 14:26:59 Checkpoint Completed: duration was 5 seconds. > > 14:26:59 Checkpoint loguniq 8922, logpos 0x2066018, timestamp:> > 0xa439eb15 > > > > 14:26:59 Maximum server connections 1> > > > > > > > The time difference between the system clocks is on both boxes is > > unnoticeable (less than a second). > > > > It seems as though the checkpoint on the Primary isn't waiting for > > the checkpoint to complete on the HDR Secondary before completing, > > which I thought it should: > > There's a bit of confusion about what we mean when we say "SYNC mode" in > context with HDR. > > Basically what that means is that the flushing of the log buffer to disk > is not considered complete (and available for reuse) until the log > buffer has been shipped to the secondary server AND the shipment of the > transmission has been ACKed. It doesn't mean that the log records have > been processed on the secondary. > > > > > " > > ... > > Checkpoints Between Database Servers > > > > Checkpoints between database servers in a replication pair are > > synchronous, regardless of the value of DRINTERVAL. (See > > "Checkpoints" on page 16-7.) A checkpoint on the primary database > > server completes only after it completes on the secondary database > > server. > > ... > > " > > > > (From "Version 10, IBM Informix Dynamic Server Administrator's > > Guide, G251-2267-02" , page 20-10) > > > > Could this be a new feature to keep checkpoints short on an HDR > > Primary? Then I'd guess this would be a documentation bug. > > > > Any thoughts? > > > > RoB > >