uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689700778
I backported this also to 8.x, removing the Ant targets related to this.
Much cleaner now!
This is an au
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689649444
@dweiss Are you fine with this? I would like to commit and backport this
soon.
Unless you have some minor comments about the cleanup process, as I had to
remove yo
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689584918
OK, reverted most of the changes in the stress tester. I kept the changes
done by @dweiss about the magic numbers like starting gun "43" or
access/release in the wire pro
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689574638
I was talking with Robert. Let's revert the in-process stuff. Makes not much
difference. We actually have a very good test for in-process logs, part of
BaseLockFactoryTes
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689563876
Code is here:
```java
try (final Lock l = verifyLF.obtainLock(lockDir, LOCK_FILE_NAME)) {
if (rnd.nextInt(10) == 0) {
if (rnd.next
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689562340
Hi again,
after reading the code of the LockStressTest: It tries to re-obtain a lock
in the same JVM based on randomization. So actually the original test did
everythi
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689546656
Hi,
I refactored the code further and removed the randomization:
- If uses `(clientId % 4)==0` to produce a thread client and otherwise a
process client
- In defa
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689480950
FYI, the limit to 2 clients was caused by ANT's xml. Every client used
another XML copypaste. But actually the number of clients is not really a
performance factor.
---
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689479947
Hi, I reordered the methods a bit in the class and applied some changes with
number of clients:
- Default is now 3. Actually this does not change the test runtime (on m
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689467186
I'd also change the test name now to `TestStressLockFactories`.
This is an automated message from the Apa
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689463433
Hi @dweiss
indeed this looks now very complicated (before the test was so simple after
my changes from last night). Indeed I want to remove some abstraction, the
fact
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689410718
> I don't want to block the changes here, but just think we should make a
followup to implement the same test, but with clients in the same JVM.
>
> There are two se
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689369526
I merged from master to retrigger the check. Clicking on "Rerun checks" did
not work, as it merged into same master commit as before. Makes no sense to me
(maybe because
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689352137
The import error was fixed already. It was not caused by this PR, but a
change in master.
This is an aut
uschindler commented on pull request #1842:
URL: https://github.com/apache/lucene-solr/pull/1842#issuecomment-689117698
I pushed some fixes:
- Removed the AtomicInteger for the port number. As the server is running in
test, we just call the server directly and it uses a Consumer(InetAddr
15 matches
Mail list logo