Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Revert "GEODE-9139: SSLException in starting up a Locator" #6635

Closed

Conversation

onichols-pivotal
Copy link
Contributor

This reverts the merge of #6436 (this fix was only intended for 1.14 and is causing problems in 1.13)

bschuchardt and others added 30 commits May 28, 2020 14:44
one more test change

(cherry picked from commit 42c1164)
* run benchmarks in parallel with other CI jobs
* eliminate JDK11 flavors for deterministic already-required-in-PR-pipeline jobs UnitTest and ApiChecker
* add Benchmarks tab

(cherry picked from commit f65ea45)
This reverts the revert of the original commit and also fixes the
originally introduced issues.

----

The OpLog initialization now delete unused drf files to prevent the
proliferation of unused records and files within the system, which
can could cause members to fail during startup while recovering
disk-stores (especially when they are isolated for gateway-senders).

- Added distributed tests.
- Delete orphaned drf files when deleting the corresponding
  crf during recovery.

(cherry picked from commit bc0090d)
* Built cleanly in multistage build
* Moved from python2 -> python3
* Pin concourse-pipeline resource

Authored-by: Robert Houghton <rhoughton@pivotal.io>
(cherry picked from commit f2392d0)
…its (apache#5232)

Remove invocation of removeEndpoint when a shared/unordered connection
shuts down.  Endpoint cleanup is already initiated by DistributionImpl
during membership view installation, so it isn't needed here.

(cherry picked from commit ece3a5a)
…ceptance tests (apache#5274)

The old keystores have expired.  I've generated 100 year keystores and added a
program to recreate them if necessary.

(cherry picked from commit a6640d7)
(cherry picked from commit 84ab66b)

Co-authored-by: Jinmei Liao <jiliao@pivotal.io>
Co-authored-by: Mark Hanson <mhanson@pivotal.io>

* Reimplement the RestoreRedundancyCommand using OperationPerformer
…pache#5257)

Co-authoried-by: Alberto Gomez <alberto.gomez@est.tech>
* During shutdown operations proxyID can be null.
	* Executing operations using this null value can cause NPEs
	* The solution is to skip all operations if the proxyID value is null.

(cherry picked from commit 7609cfa)
…ation is disabled (apache#5250)

* GEODE-8144: endpoint identification in servers is not working

modified the fix for this issue to not set the SNI server name parameter
if endpoint verification is disabled.  We're doing this because setting
this parameter appears to decrease performance in large performance
tests.

* changed test to throw exceptions instead of asserting they don't exist

* replaced check for SNI server name in SSL parameters with a more in-depth check

* SSLParameters.getServerNames() may return a null value

(cherry picked from commit b1107d2)
Authored-by: Sean Goller <sgoller@pivotal.io>
(cherry picked from commit 15d6d26)
…tenerImpl (apache#5306)

I've replaced the "for" loop using an implicit Iterator with one using an
explicit Iterator so that its safe "remove()" method can be used.  The
Iterator method is stated as being the only safe way to modify the
collection while iterating over its contents.

I've also modified a test to validate the fix.  The test forces a
failure to send two messages to an address.  The failures are then
handled in the code that was throwing the
ConcurrentModificationException and, since there are two failures,
it causes two removals to be performedon the failedMessages collection.

(cherry picked from commit 3cda1b1)
(cherry picked from commit 5732a1c)

Co-authored-by: Mark Hanson <mhanson@pivotal.io>
…e#5308)

Make it clearer that neither a local Bookbinder nor Ruby interpreter is
needed to build our documentation book.
…Command (apache#5300) (apache#5327)

* GEODE-8095: Changes to make GEODE Respond to Restore Redundancy

(cherry picked from commit f5c5e2c)

Co-authored-by: Jinmei Liao <jiliao@vmware.com>
…eption, it should retry (apache#5253)

    Co-authored-by: Xiaojian Zhou <gzhou@pivotal.io>
    Co-authored-by: Anil <agingade@pivotal.io>

(cherry picked from commit ee9a4b0)
)

Do not delete drf files during member startup as that should be only
done by the compactor thread. Instead, allow the OplogEntryIdSet to
grow over the default capacity and log a warning message instructing
the user to manually compact the disk-stores.

- Added unit tests.
- Replaced usages of 'junit.Assert' by 'assertj'.
- Modified DiskStoreImpl.deadRecordCount to return long instead of int.
- Added internal overflow implementation to the OplogEntryIdSet so it can
  grow above the default limit.

(cherry picked from commit fdc4401)
* GEODE-8020: buffer management problems

This fixes some buffer handling in MsgStreamerList and alters
MstStreamer to avoid creating MsgStreamerList and VersionedMsgStreamers
during normal, non-upgrade, operations.

It also changes NioSslEngine to use synchronization in more places,
notably the close() method, which was possibly allowing multiple threads to
change the state of the engine.

* revert unnecessary change to ClusterCommunicationsDUnitTest

* fixing another null version check

* renamed new BufferPool property

* restore logging of ssl exceptions

(cherry picked from commit 7375c59)
…the way we release (apache#5361)

also balance max_in_flight according to how long each job takes

(cherry picked from commit daa70d7)
…che#5396)

* GEODE-8377: User Guide: GFSH GC command should be documented to use only in non-production environments
Co-authored-by: Karen Miller <kmiller@pivotal.io>
kirklund and others added 8 commits June 11, 2021 09:58
Reasons for changing this target to non-required:
* stress-new-test stresses existing unchanged tests in addition to new and updated tests
* we now have codeowners and weekly mass test runs, which provide better feedback
* most of the other PR checks are not required and act as advisory input to reviewers

(cherry picked from commit 7a886b7)
…ssage

	* Even if there any any exception, a reply will be sent back to the sender so that the sender's threads are not stuck.

(cherry picked from commit 31bb9b9)
(cherry picked from commit c9d4f68)
This reverts commit 35cbee2.
…tTest.testReconnectAfterMeltdown hang (apache#5192)"

This reverts commit aba29fb.
This reverts commit 844f2a7.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet