Github user jvrao commented on the issue:
https://github.com/apache/bookkeeper/pull/189
@sijie @reddycharan Let me step back and discuss this.
1. Are there any legitimate use cases of specifying multiple ledger dirs on
the same disk partition? if not, in production I would like to flag this as
error, or at least a warning to ignore other ledger dirs. I don't think code
need to work hard to dedup and skip these things. This is not a naive user
mistake, configurations are done by system administrator and they need to be
crisp IMO.
2. Currently we have only one active entrylog, so having multiple dirs on
the same disk, at the most may fill one disk partition faster/quicker than
other. (Say three ledger dirs on disk1 and one ledger dir on disk2). If disk1
and disk2 are of same sizes, disk1 gets used more. but if disk1 is bigger than
disk2 then this feature may be handy.
3. When we move to multiple active entrylogs, we are going to distribute
entrylogs across ledger dirs. When this feature is introduced it surely hits
perf because one disk may be having multiple entrylogs which gets
written/scyned unless dedup is implemented.
@sijie I am not sure what is the biggest concern about this. Are you
worried about too many configuration parameters? or something else?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
