Thanks Stephen On Wed, Dec 21, 2022 at 11:54 PM Stephen Darlington < stephen.darling...@gridgain.com> wrote:
> Please note that volunteers answer questions here. If you need SLAs for > answers to your questions, there are companies that provide support. > > Full sync would work. You could also use transactions. > > On 21 Dec 2022, at 17:58, Surinder Mehra <redni...@gmail.com> wrote: > > Gentle reminder. > > On Tue, 20 Dec 2022, 12:53 Surinder Mehra, <redni...@gmail.com> wrote: > >> Hi, >> If we have cache synchronization mode as Primary Sync and read from >> backup as true(default), would this cause consistency issues if primary >> goes down without flushing all data to disk. >> I understand WAL is used to replay unflushed data when a node comes back >> online again. But if say out of 10 records written to cache, only 5 of them >> were written to disk by checkpointing thread and data was not replicated to >> the backup node yet(PRIMARY-SYNC) when primary node went node, the client >> would assume backup node would have all 10 records . But it might not be >> the case. >> >> So am I correct in assuming we need to either turn off read from backup >> or change cache synchronization mode to FULL-SYNC >> >> Answer to stackoverflow question below suggests disabling read from >> backup. Do we need to disable it even in FULL-SYNC mode ? >> >> >> https://stackoverflow.com/questions/67598175/ignite-read-stale-data-from-backup-node >> > >