[
https://issues.apache.org/jira/browse/HDFS-1936?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13034446#comment-13034446
]
Todd Lipcon commented on HDFS-1936:
-----------------------------------
I compiled the following list of versions:
{noformat}
// -35: Adding support for block pools and multiple namenodes (federation)
// -34: no new feature - just bumped after following three were reserved
// The following three image versions are special out-of-sequence allocations
// (i.e. they don't necessarily contain all features from "earlier" versions)
// See HDFS-1822, HDFS-1842,
// -33: 0.22.0
// -32: 0.20.204
// -31: 0.20.203
// -30: HDFS-1070: store only last component of a path in image
// -29: unused (skipped for no reason)
// -28: HDFS-1630: fsedits are checksummed
// -27: HDFS-259: remove intentionally corrupt pre-0.13 image directory
// -26: image checksum
// -25: image compression
// -24: added OP_*_DELEGATION_TOKEN and OP_UPDATE_MASTER_KEY *** 0.21.0 ***
// -23: symlinks
// -22: OP_CONCAT_DELETE
// -21: atomic rename function
// -20: datanode has a "rbw" subdirectory (append impl in 0.21)
// -19: sticky bit *** CDH3, earlier 0.20.203 ***
// -18: support disk space quotas *** 0.20.2 ***
// -17: support access time on files
// -16: change edit log and fsimage to support quotas
// -15: store generation stamp with each block
// The special versions -31, -32, -33 are as follows:
// -31: corresponds to release 0.20.203:
// * Includes functionality up to version -19
// * includes delegation token opcodes from -24, BUT
// with different opcode identifiers. Given this,
// we don't allow upgrade from this version.
// -32:
// * Includes same set of features as version -31
// * Distinction is that the opcodes match trunk
//
// -33: 0.22.0
// * contains all features up through version -27
{noformat}
It seems that this patch might break upgrade from 0.21?
Rather than trying to force all the checks to particular "release boundaries",
I really think it would be easier to follow if we did the refactor now so that
we can just encode that compression is in -25 through -30 and -35 onwards?
> Updating the layout version from HDFS-1822 causes upgrade problems.
> -------------------------------------------------------------------
>
> Key: HDFS-1936
> URL: https://issues.apache.org/jira/browse/HDFS-1936
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: name-node
> Affects Versions: 0.22.0, 0.23.0
> Reporter: Suresh Srinivas
> Assignee: Suresh Srinivas
> Priority: Blocker
> Fix For: 0.22.0, 0.23.0
>
> Attachments: HDFS-1936.22.patch, HDFS-1936.trunk.patch
>
>
> In HDFS-1822 and HDFS-1842, the layout versions for 203, 204, 22 and trunk
> were changed. Some of the namenode logic that depends on layout version is
> broken because of this. Read the comment for more description.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
