[
https://issues.apache.org/jira/browse/OAK-12128?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18066334#comment-18066334
]
Stefan Egli commented on OAK-12128:
-----------------------------------
bq. Q1: does this make sense at all?
yes I think it does
bq. Q2: exception? log? both?
I commented in the PR : I would consider logging twice, first without, then
with exception as the latter might fail due to memory pressure
bq. Q3: write operations or all?
I see you did it to all now. I guess it doesn't hurt. reads are likely executed
much more frequently, which would mean we might get more frequent logs, which
would mean this way we have a higher chance of eventually getting an idea how
frequently this happens out there.
bq. Q4: add more details to exception? (like method invoked - however stack
trace will tell us anyway)
looks fine the way it is now
> Fail earlier (and add more diags) when detecting late writes
> ------------------------------------------------------------
>
> Key: OAK-12128
> URL: https://issues.apache.org/jira/browse/OAK-12128
> Project: Jackrabbit Oak
> Issue Type: Task
> Components: documentmk
> Reporter: Julian Reschke
> Assignee: Julian Reschke
> Priority: Major
> Attachments: leasecheck.diff
>
>
> While discussing late write with [~stefanegli] we realized that late writes
> (or at least some of them) can actually be detected in
> LeaseCheckDocumentStoreWrapper - if a write call starts *before* lease end
> and completes *after* lease end, it is (likely?) a late write and we can
> either make it fail, or at least log it.
> Q1: does this make sense at all?
> Q2: exception? log? both?
> Q3: write operations or all?
> Q4: add more details to exception? (like method invoked - however stack trace
> will tell us anyway)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
