[
https://issues.apache.org/jira/browse/HDFS-1801?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13030563#comment-13030563
]
Todd Lipcon commented on HDFS-1801:
-----------------------------------
bq. IIUC the sequence in your comment above only applies when the edits dir
lives on a different volume from the image dir.
Yes, that's correct. But with the 1073 changes, there is no longer such tight
coupling between image state and edits state, and that coupling gets even
looser once we make the edits storage more pluggable.
bq. Doesn't the last test in TestNameEditsConfigs#testNameEditsConfigsFailure
need an fail("...") before the catch since we expect this last test to fail? Ie
if it does not fail the test can still pass.
hrm, not sure which test you're referring to. The very last one is commented
out. The second to last one has an assertTrue(false) like it used to (no change
there)
bq. Header comment of readTransactionIdFile still refers to time
Fixed and also elaborated in the javadoc.
bq. Useful to have a comment by NameNodeFile.TIME indicating this is
pre-HDFS-1073
Fixed
bq. Would CKPT_TXID be a more descriptive name than SEEN_TXID since this is the
txid at the time of checkpoint?
I agree that SEEN_TXID is a kind of bad name, but checkpoint doesn't describe
it either. It's also incremented on every log roll.
bq. The removed edit log ops codes need to be reintroduced in the future to fix
the BN right? In which case perhaps leave them in for now.
Actually, I don't think they will need to be introduced - the new
start/end-logsegment ops should be sufficient.
> Remove use of timestamps to identify checkpoints and logs
> ---------------------------------------------------------
>
> Key: HDFS-1801
> URL: https://issues.apache.org/jira/browse/HDFS-1801
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: name-node
> Affects Versions: Edit log branch (HDFS-1073)
> Reporter: Todd Lipcon
> Assignee: Todd Lipcon
> Fix For: Edit log branch (HDFS-1073)
>
> Attachments: hdfs-1801.txt, hdfs-1801.txt, hdfs-1801.txt,
> hdfs-1801.txt
>
>
> Currently, the NameNode validates checkpoint uploads by using timestamps
> associated with checkpoints and edit logs. However, now that we have
> transaction IDs that uniquely identify each point in time in the history of a
> namespace, it is more robust to simply use transaction IDs to identify images
> and edits.
> This JIRA is to remove the use of editsTime and checkpointTime and replace it
> with:
> * {{lastCheckpointTxId}} - the highest transaction ID reflected in the most
> recently saved fsimage file
> * {{lastLogRollTxId}} - the highest transaction ID in {{edits}} when
> {{rollFsImage}} was called by the checkpointing node.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
