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
Upgrade to JRuby 9k #5606
Comments
I can revamp #3654 in more structured way, including the intention to clarify possible fast/slow areas due to the change to 9k, so we can work on them. |
Found this on the date filter with 9.1.5.0
same test passes in 1.7.25 |
I have begun the upgrade process. All core specs pass with JRuby 9.1.6.0 at the time of posting this comment. I will be adding notes here about what needs to change in the plugins as I find them (as well as adding an issue to each plugin) |
|
Netflow Codec, JRuby 1.7.X quirk. |
CSV Output I have implemented the method in a thread-safe way locally. |
RabbitMQ output. double received |
|
@guyboertje the rabbitmq bug was fixed in logstash-plugins/logstash-output-rabbitmq#56 |
@guyboertje #6576 is also a blocker. |
lsperfm 20 second run 9K
1.7.25
|
Once we have a PR to move logstash into jruby 9k we should also start testing against java 9, so we can help validate the jruby support (see jruby/jruby#4111) |
New longer lsperfm benchmarks with % change. Averaged 3 x 120 second runs. The data in the table that was below is erroneous. Removed table. [table removed] |
longer lsperfm benchmarks with % change, 6 minutes each run Please scroll to view the right most columns. In the configs that include a filter the difference between 1.7.25 and 9K range from -1% to +3% - this is a good sign. The data in the table that was below is erroneous. Removed. Comment below refers to the erroneous data. [table removed] |
@guyboertje Any ideas explaining the tremendous discrepancies between 1.7.25 & 9K with the |
I will rerun the tests. On the 1.7.25 instance the log.level was set to warn. I will set both instances to warn. NOTE: my current JVM is Azul Zulu for all benchmarks above. |
Please scroll to view the right most columns. The difference between 1.7.25 and 9K range from -4% to +5% - this is a good sign as there are no wildly differing numbers across the suite. Further -4% to +5% could be caused by interruptions from other processes on my laptop. To achieve this, I warmed up the JVM by running one 90 run first then the normal suite for 6 minutes each. Both instances of Logstash had their logging level set to
|
Now i'm using logstash 5.2 with To use output-csv plugin with logstash 5.x , the method received need to be remplaced by multi_receive_encoded method #11 suggested solution (edit csv.rb). but it seems data can be lost compare to the original one (the broken one) and also some of csv option can't be use either(such as add csv header) . So could you please suggest the solution for writing as csv file with logstash 5.x? Thank you! |
@gunpix |
@guyboertje do you have a branch for this work? I'm running into issues on Jruby 9k (testing on Java 9 for #6891). |
@jordansissel - I have this PR #6713. |
Adding a few traces I've seen in the bootrap / running test. I will just list a few issue I've seen in the build to make
|
With the update of rspec we also see this in the suite, we will need to be more robust in the expect when using raise_error
|
First run doesn't show too much errors:
|
Nothing really major, JRuby 9k seems to have switched to java.time for the implementation of the |
@ph - I have discovered some weirdnesses with Joda versions and the date filter. I am adding a general LS issue to cover a move from Joda to Java 8 Time across the whole of LS. |
@guyboertje Yup, I believe this move can be done after we move to JRuby 9K. The case that come to my mind is the following:
|
Everything is passing locally with the master branch, the move was pretty smooth, I am running the test of all the plugins against the core and will update this issue with the changes. |
qa/integration passes, I am running into issue for the ruby 1.7.25
jruby 9k
My theory is there is a bug in fpm and we close the pipe/channel at the wrong time. |
We will need to update to http://jruby.org/2017/05/25/jruby-9-1-10-0 |
Also when building plugins we see this warning:
|
Previous error only seems to be on plugins build. |
I've found another regression, 9k
master
5.4.0
EDIT: fixed by making sure we implement inspect on Timestamp. |
#7349 is merged |
According to this tweet, JRuby 1.7 will be EOL by the sometime during 2016.
Let's make sure we have all the bugs/problems identified and get on the path to upgrade to JRuby 9k some time after Logstash 5.0.0
The text was updated successfully, but these errors were encountered: