You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-dev@jackrabbit.apache.org by Michael Dürig <md...@apache.org> on 2016/04/19 10:24:19 UTC
Heads up, humps ahead: merge of big patch for OAK-3348
Hi,
After weeks of testing we are finally in a position to go forward with a
fix for OAK-3348 [1]. This is a prerequisite for the segment store
revision gc.
The amount of changes required for this fix is quite big and there will
still be a couple of loose ends to knot once applied. Most prominently
there will be incompatible changes in the persistence format: previous
segment stores *WILL NOT WORK* once this fix is applied. However fear
not, we will come up with clean migration paths as part of this effort.
So please bear with us in the next couple of weeks should it be a bit
bumpy. However, apart from the format change none of the changes should
directly affect any module outside of oak-segment. Should at some point
oak-segment be too unstable for your purposes I suggest to temporarily
replace the snapshot dependency to oak-segment with the latest release
version.
Please review the fix for OAK-3348 at the respective GitHub branch [2]
if you want to get your hand dirty. I placed FIXME OAK-3348 tags where
further work is required and will follow up with dedicated issues as
needed. Specifically there is a couple of tests that need fixing (mostly
related to the change in persistent format and most of them within
oak-segment):
oak-jcr:
UpgradeTest.upgradeFrom10
oak-upgrade:
AbstractOak2OakTest.validateMigration
RepositorySidegradeTest.verifyGenericProperties
oak-segment:
FileStoreBackupTest
CompactionAndCleanupIT.compactionNoBinaryClone
CompactionAndCleanupIT.propertyRetention
SegmentGraphTest
SegmentIdFactoryTest.dataAIOOBE
SegmentNodeStoreServiceTest
SegmentSizeTest
I disabled them for now until sorted out.
Michael
[1] https://issues.apache.org/jira/browse/OAK-3348
[2] https://github.com/mduerig/jackrabbit-oak/commits/OAK-3348
Re: Heads up, humps ahead: merge of big patch for OAK-3348
Posted by Michael Dürig <md...@apache.org>.
Hi,
The main parts are in place by now. The new module oak-segment-next
basically replaces oak-segment:
- all commits should go to oak-segment-next, we should not touch
oak-segment any more. Please ensure to rework any patches you might have
pending.
- most of the issues for oak-segment moved to oak-segment-next. There is
only a couple left for the former.
- we are still working out the details of when to switch the default
from oak-segment to oak-segment-next, migration and releasing [1]. Will
follow up on this one once there.
Michael
[1] https://issues.apache.org/jira/browse/OAK-4243
On 20.4.16 10:49 , Michael Dürig wrote:
>
> Hi,
>
> To follow up on this, there was a change in plan. As applying the
> changes on top of the current oak-segment module would make migration
> very difficult later, we need to introduce a new module. See OAK-4243
> for details on this [1].
>
> This voids most of my heads up from below as we can stabilise our
> implementation in isolation and switch to it once we consider it ready.
>
> Michael
>
> [1] https://issues.apache.org/jira/browse/OAK-4243
>
>
> On 19.4.16 10:24 , Michael Dürig wrote:
>>
>> Hi,
>>
>> After weeks of testing we are finally in a position to go forward with a
>> fix for OAK-3348 [1]. This is a prerequisite for the segment store
>> revision gc.
>>
>> The amount of changes required for this fix is quite big and there will
>> still be a couple of loose ends to knot once applied. Most prominently
>> there will be incompatible changes in the persistence format: previous
>> segment stores *WILL NOT WORK* once this fix is applied. However fear
>> not, we will come up with clean migration paths as part of this effort.
>>
>> So please bear with us in the next couple of weeks should it be a bit
>> bumpy. However, apart from the format change none of the changes should
>> directly affect any module outside of oak-segment. Should at some point
>> oak-segment be too unstable for your purposes I suggest to temporarily
>> replace the snapshot dependency to oak-segment with the latest release
>> version.
>>
>> Please review the fix for OAK-3348 at the respective GitHub branch [2]
>> if you want to get your hand dirty. I placed FIXME OAK-3348 tags where
>> further work is required and will follow up with dedicated issues as
>> needed. Specifically there is a couple of tests that need fixing (mostly
>> related to the change in persistent format and most of them within
>> oak-segment):
>>
>> oak-jcr:
>> UpgradeTest.upgradeFrom10
>>
>> oak-upgrade:
>> AbstractOak2OakTest.validateMigration
>> RepositorySidegradeTest.verifyGenericProperties
>>
>> oak-segment:
>> FileStoreBackupTest
>> CompactionAndCleanupIT.compactionNoBinaryClone
>> CompactionAndCleanupIT.propertyRetention
>> SegmentGraphTest
>> SegmentIdFactoryTest.dataAIOOBE
>> SegmentNodeStoreServiceTest
>> SegmentSizeTest
>>
>> I disabled them for now until sorted out.
>>
>> Michael
>>
>> [1] https://issues.apache.org/jira/browse/OAK-3348
>> [2] https://github.com/mduerig/jackrabbit-oak/commits/OAK-3348
>>
Re: Heads up, humps ahead: merge of big patch for OAK-3348
Posted by Michael Dürig <md...@apache.org>.
Hi,
To follow up on this, there was a change in plan. As applying the
changes on top of the current oak-segment module would make migration
very difficult later, we need to introduce a new module. See OAK-4243
for details on this [1].
This voids most of my heads up from below as we can stabilise our
implementation in isolation and switch to it once we consider it ready.
Michael
[1] https://issues.apache.org/jira/browse/OAK-4243
On 19.4.16 10:24 , Michael Dürig wrote:
>
> Hi,
>
> After weeks of testing we are finally in a position to go forward with a
> fix for OAK-3348 [1]. This is a prerequisite for the segment store
> revision gc.
>
> The amount of changes required for this fix is quite big and there will
> still be a couple of loose ends to knot once applied. Most prominently
> there will be incompatible changes in the persistence format: previous
> segment stores *WILL NOT WORK* once this fix is applied. However fear
> not, we will come up with clean migration paths as part of this effort.
>
> So please bear with us in the next couple of weeks should it be a bit
> bumpy. However, apart from the format change none of the changes should
> directly affect any module outside of oak-segment. Should at some point
> oak-segment be too unstable for your purposes I suggest to temporarily
> replace the snapshot dependency to oak-segment with the latest release
> version.
>
> Please review the fix for OAK-3348 at the respective GitHub branch [2]
> if you want to get your hand dirty. I placed FIXME OAK-3348 tags where
> further work is required and will follow up with dedicated issues as
> needed. Specifically there is a couple of tests that need fixing (mostly
> related to the change in persistent format and most of them within
> oak-segment):
>
> oak-jcr:
> UpgradeTest.upgradeFrom10
>
> oak-upgrade:
> AbstractOak2OakTest.validateMigration
> RepositorySidegradeTest.verifyGenericProperties
>
> oak-segment:
> FileStoreBackupTest
> CompactionAndCleanupIT.compactionNoBinaryClone
> CompactionAndCleanupIT.propertyRetention
> SegmentGraphTest
> SegmentIdFactoryTest.dataAIOOBE
> SegmentNodeStoreServiceTest
> SegmentSizeTest
>
> I disabled them for now until sorted out.
>
> Michael
>
> [1] https://issues.apache.org/jira/browse/OAK-3348
> [2] https://github.com/mduerig/jackrabbit-oak/commits/OAK-3348
>