Tar file naming conventions

[marcush35887118]

The tar files used by AEM in the repository/segmentstore have names like dataNNNNNL.tar, where NNNNN is a 5 digit number and L is a letter. What happens when NNNNN needs to be greater than 99999?

[arunpatidar · Accepted solution]

Hi,

Can you please check below if it helps:

Oak TAR file layout

A few words on how Oak names TAR files. The convention is to always start with a data00000a.tar file. As data is written to the repository, new TAR files are added with increasing numbers, thus ending up with data00001a.tar, data00002a.tar and so on.

Each time a compaction cycle ends, there is a cleanup phase in which segments from an old generation are purged. Those tar files that shrink by at least 25% are rewritten to a new tar generation, skipping the reclaimed segments. A shrunk TAR file increases its tail generation character, e.g. from data00000a.tar to data00000b.tar.

More info ☞ 

[marcush35887118]

Thanks for the info and link. Our webmaster runs compaction weekly and was noticing the numbers steadily increasing and was wondering how the naming conventions would be handled since we're up around 64832.

[aemmarc]

What happens when NNNNN needs to be greater than 99999?

Better run Compaction before the World comes to an end if that is the case. 

99,999 TAR files at 256 MB per TAR is an utterly ridiculous 25 TB segmentstore size.

[marcush35887118]

Of course this comment wasn't helpful and it's not even remotely as funny as you thought it was. Do better next time.