When you manually force a rollup or the core performs the Nightly Job which includes the rollup process which processes the retention policies for a protected agent you may receive errors about Missing EPOCHS and the job fails with an error similar to the following:
Epoch chain broken for volume 'C:\Windows\SRPPartition'. Missing epoch='2329'. Ignoring bad epoch chain from possible recovery points '2509' to epoch '2376'
If the core was performing a rollup for an agent and that process was interrupted due to a server issue or disk related issue then the file the rollup was processing may become damaged. This is similar to when Windows is writing to a file and an issue arises that causes that write to fail which can result in bad blocks.
Another cause is if an agent was transferring a snapshot and the core was writing the EPS file to the repository but the transfer was interrupted due to network timeout, agent vss issue or core processing issue the file may be left behind. This can cause invalid epoch errors and the rollup may not be able to process the recovery point file because it is incomplete.
Adding the following registry entry to the core registry for the specific agent that failed the rollup the core can remove these invalid files and correct recovery point chain issues. This can also result in additional free space in the repository depending on how large those files were that the core removed.
***In the step below Director is just an example for one of my protected servers but you will be looking for your agent server names***
1) Open Regedit and navigate to HKLM\SOFTWARE\Appassure\ReplayEPS\Director
2) Create a new dword value for cleanup_invalid_files
3) Set the decimal value to 1
4) Reboot the core that the reg edits were just performed on
Allow the core to process the nightly job so that retention is applied and monitor the events to verify if the EPOCH chain errors persist