How are you doing backups? 1) running nodetool snapshot command on all nodes at once 2) while cluster is backup mode we take the EBS snapshot backup of that volume 3) running the nodetool clearsnapshot for existing the backup mode.
How are you doing restores? 1) Creating volume from snapshot 2) Attaching those volume to DR instances 3) in cassandra.yaml we are passing the token range for each instance. What consistency level are you using for writes? Reads? In DR we are using QUORUM consistency for Reads Is the data in the sstable (can you find it with sstabledump?) can you provide us the steps to for the same On Mon, Dec 2, 2024 at 12:05 PM Jeff Jirsa <jji...@gmail.com> wrote: > 3.11.2 is from Feb 2018. It’s a 6 year old release. > > It’s VERY hard to guess what’s happening here without a lot more info. > > How are you doing backups? > How are you doing restores? > What consistency level are you using for writes? Reads? > Is the data in the sstable (can you find it with sstabledump?) > > > > On Dec 1, 2024, at 10:03 PM, guo Maxwell <cclive1...@gmail.com> wrote: > > Hello, > You mean 3.11.2 will have the problem of backup data loss, but 3.11.4 will > not have this problem, right? > > Can you provide the backup and recovery steps for these two versions? If > possible, it would be better to give a few examples. > For example, if three data a, b, and c are inserted, and a will be lost in > 3.11.2 but this situation will not happen again in 3.11.4. > > I believe more information can help us better understand and solve this > problem. > > Soyal Badkur <soyalbad...@gmail.com> 于2024年12月2日周一 13:43写道： > >> Hi Team, >> >> We are facing issues while restoring Cassandra backup from PROD servers >> to DR servers for Cassandra Version 3.11.2. When we try to read the data >> from both the environments to verify the data backups, we found that in DR >> environment data is not showing as accepted. We have to run the same query >> multiple times in CQLSH to find that data is present in the DR side but >> unable to do so while running the script. But in some cases we find that >> some of the data is lost in DR even though it is present in PROD, without >> any delete query run on PROD side or its TTL is expired(basically we are >> running the script from n-2 days to compare the data). >> >> When we do the same for another cluster which is in 3.11.4 Cassandra >> version we don't face any such issues. Please help us in fixing this issue. >> >> Do let us know if you need any more information from us. >> >> Regards, >> Soyal Badkur >> > >
