You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@metron.apache.org by justinleet <gi...@git.apache.org> on 2017/09/08 19:38:50 UTC
[GitHub] metron pull request #741: METRON-1153 HDFS HdfsWriter never recovers from ex...
GitHub user justinleet opened a pull request:
https://github.com/apache/metron/pull/741
METRON-1153 HDFS HdfsWriter never recovers from exceptions
## Contributor Comments
Added a try-catch around the actual write that will rotate the file and try again if there's a stream closed underneath it. Added two unit tests, one that ensures things flow through nicely to a single file with a double write and one that ensures that things flow to two files if the channel is closed underneath for whatever reason (done by just calling `closeOutputFile()` when outside of the normal flow).
It's not a perfect solution, but it should alleviate any transient pain and let us know if the problem is deeper if it keeps showing up.
Also added a missing set to a field from the constructor which probably wasn't helping things since I happened to notice it. It's a one line change, so it seemed excessive to create a separate PR.
## Pull Request Checklist
Thank you for submitting a contribution to Apache Metron.
Please refer to our [Development Guidelines](https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=61332235) for the complete guide to follow for contributions.
Please refer also to our [Build Verification Guidelines](https://cwiki.apache.org/confluence/display/METRON/Verifying+Builds?show-miniview) for complete smoke testing guides.
In order to streamline the review of the contribution we ask you follow these guidelines and ask you to double check the following:
### For all changes:
- [x] Is there a JIRA ticket associated with this PR? If not one needs to be created at [Metron Jira](https://issues.apache.org/jira/browse/METRON/?selectedTab=com.atlassian.jira.jira-projects-plugin:summary-panel).
- [x] Does your PR title start with METRON-XXXX where XXXX is the JIRA number you are trying to resolve? Pay particular attention to the hyphen "-" character.
- [x] Has your PR been rebased against the latest commit within the target branch (typically master)?
### For code changes:
- [x] Have you included steps to reproduce the behavior or problem that is being changed or addressed?
- [x] Have you included steps or a guide to how the change may be verified and tested manually?
- [x] Have you ensured that the full suite of tests and checks have been executed in the root metron folder via:
```
mvn -q clean integration-test install && build_utils/verify_licenses.sh
```
- [x] Have you written or updated unit tests and or integration tests to verify your changes?
- [x] If adding new dependencies to the code, are these dependencies licensed in a way that is compatible for inclusion under [ASF 2.0](http://www.apache.org/legal/resolved.html#category-a)?
- [x] Have you verified the basic functionality of the build by building and running locally with Vagrant full-dev environment or the equivalent?
### For documentation related changes:
- [x] Have you ensured that format looks appropriate for the output in which it is rendered by building and verifying the site-book? If not then run the following commands and the verify changes via `site-book/target/site/index.html`:
```
cd site-book
mvn site
```
#### Note:
Please ensure that once the PR is submitted, you check travis-ci for build issues and submit an update to your PR as soon as possible.
It is also recommended that [travis-ci](https://travis-ci.org) is set up for your personal repository such that your branches are built there before submitting a pull request.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/justinleet/metron METRON-1153
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/metron/pull/741.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #741
----
commit e5a2f3c6114f26b2091640a57bf4ca02e74addfa
Author: justinjleet <ju...@gmail.com>
Date: 2017-09-08T19:34:08Z
adding attempt to get new file on channel closed exception
----
---
[GitHub] metron pull request #741: METRON-1153 HDFS HdfsWriter never recovers from ex...
Posted by ottobackwards <gi...@git.apache.org>.
Github user ottobackwards commented on a diff in the pull request:
https://github.com/apache/metron/pull/741#discussion_r137913406
--- Diff: metron-platform/metron-writer/src/main/java/org/apache/metron/writer/hdfs/SourceHandler.java ---
@@ -64,19 +64,34 @@ public SourceHandler(List<RotationAction> rotationActions
this.rotationPolicy = rotationPolicy;
this.syncPolicy = syncPolicy;
this.fileNameFormat = fileNameFormat;
+ this.cleanupCallback = cleanupCallback;
initialize();
}
protected void handle(JSONObject message, String sensor, WriterConfiguration config, SyncPolicyCreator syncPolicyCreator) throws IOException {
byte[] bytes = (message.toJSONString() + "\n").getBytes();
synchronized (this.writeLock) {
- out.write(bytes);
+ try {
+ out.write(bytes);
+ } catch (IOException writeException) {
+ LOG.warn("IOException while writing output", writeException);
+ // If the stream is closed, attempt to rotate the file and try again, hoping it's transient
+ if (writeException.getMessage().contains("Stream Closed")) {
--- End diff --
Using string compare on the message seems kind of thin. And there may be other exceptions that come up that we want to handle.
Part of what I hoped would come from the discuss thread was a discussion about what was recoverable and what was not recoverable, and a better way to handle both cases.
---
[GitHub] metron pull request #741: METRON-1153 HDFS HdfsWriter never recovers from ex...
Posted by asfgit <gi...@git.apache.org>.
Github user asfgit closed the pull request at:
https://github.com/apache/metron/pull/741
---
[GitHub] metron issue #741: METRON-1153 HDFS HdfsWriter never recovers from exception...
Posted by ottobackwards <gi...@git.apache.org>.
Github user ottobackwards commented on the issue:
https://github.com/apache/metron/pull/741
@justinleet My question with regards to the necessity for a refactor came from wanting to handle the exception in the HDFSWriter, where the exceptions where being caught at the time.
The fact that the tuples are 'unpaired' from their messages, and that we handle them all together seems to me to be problematic if you want to take per message/sourcehandle action.
I think your approach removes the immediate problem with that, although having both the hdfs writer and the source handler itself take actions and split the 'ownership' doesn't feel quite right to me.
@cestella It would be good to know why this is happening, but in truth, any persistent network connection in a long lived 'forever' type application needs to guard against these kinds of errors.
For reliability reasons, and where we want to get to, we require a more clear documentation of the failure and recovery states of the writers, esp. the hdfs as we are batching. We also need to understand all the ways the stream can fail, to the extent that that is possible.
I am not sure that limiting it to this one case is enough, there will still be many possible ways to end up with a very unclear situation, that the writer is failing continuously and but the source handler is not removed from 'service'. Users, as we have seen on the list will be left to track through system to work their way to this problem.
Failure to store is a critical failure of the system. Esp. in systems where there are data retention rules or SLA's on data loss. Thus we need in addition to the handling of this, an alerting strategy.
While this fix ( questions pending ) is an improvement of the symptom, it does not address the higher level issue or severity of this problem.
---
[GitHub] metron pull request #741: METRON-1153 HDFS HdfsWriter never recovers from ex...
Posted by ottobackwards <gi...@git.apache.org>.
Github user ottobackwards commented on a diff in the pull request:
https://github.com/apache/metron/pull/741#discussion_r137913296
--- Diff: metron-platform/metron-writer/src/main/java/org/apache/metron/writer/hdfs/SourceHandler.java ---
@@ -64,19 +64,34 @@ public SourceHandler(List<RotationAction> rotationActions
this.rotationPolicy = rotationPolicy;
this.syncPolicy = syncPolicy;
this.fileNameFormat = fileNameFormat;
+ this.cleanupCallback = cleanupCallback;
initialize();
}
protected void handle(JSONObject message, String sensor, WriterConfiguration config, SyncPolicyCreator syncPolicyCreator) throws IOException {
byte[] bytes = (message.toJSONString() + "\n").getBytes();
synchronized (this.writeLock) {
- out.write(bytes);
+ try {
+ out.write(bytes);
+ } catch (IOException writeException) {
+ LOG.warn("IOException while writing output", writeException);
+ // If the stream is closed, attempt to rotate the file and try again, hoping it's transient
+ if (writeException.getMessage().contains("Stream Closed")) {
+ LOG.warn("Output Stream was closed. Attempting to rotate file and continue");
+ rotateOutputFile();
+ // If this write fails, the exception will be allowed to bubble up.
+ out.write(bytes);
--- End diff --
If there is a non-recoverable error, you are in the same position as before.
I was thinking of having a 'failed' flag ( similar to what hadoop has in it's screams where they track closed explicitly )
---
[GitHub] metron issue #741: METRON-1153 HDFS HdfsWriter never recovers from exception...
Posted by cestella <gi...@git.apache.org>.
Github user cestella commented on the issue:
https://github.com/apache/metron/pull/741
I absolutely agree, this does not address the higher level issue or severity. This is merely a stopgap to fix something that we are seeing in the field and on the list. I made a JIRA to capture this: https://issues.apache.org/jira/browse/METRON-1170
---
[GitHub] metron issue #741: METRON-1153 HDFS HdfsWriter never recovers from exception...
Posted by cestella <gi...@git.apache.org>.
Github user cestella commented on the issue:
https://github.com/apache/metron/pull/741
This looks good to me as a stopgap. I'd like to understand eventually why this happens, but it appears very intermittent, so this isn't a bad solution at least as a stopgap.
+1 by inspection.
---
[GitHub] metron pull request #741: METRON-1153 HDFS HdfsWriter never recovers from ex...
Posted by ottobackwards <gi...@git.apache.org>.
Github user ottobackwards commented on a diff in the pull request:
https://github.com/apache/metron/pull/741#discussion_r137913278
--- Diff: metron-platform/metron-writer/src/main/java/org/apache/metron/writer/hdfs/SourceHandler.java ---
@@ -64,19 +64,34 @@ public SourceHandler(List<RotationAction> rotationActions
this.rotationPolicy = rotationPolicy;
this.syncPolicy = syncPolicy;
this.fileNameFormat = fileNameFormat;
+ this.cleanupCallback = cleanupCallback;
initialize();
}
protected void handle(JSONObject message, String sensor, WriterConfiguration config, SyncPolicyCreator syncPolicyCreator) throws IOException {
byte[] bytes = (message.toJSONString() + "\n").getBytes();
synchronized (this.writeLock) {
- out.write(bytes);
+ try {
+ out.write(bytes);
+ } catch (IOException writeException) {
+ LOG.warn("IOException while writing output", writeException);
+ // If the stream is closed, attempt to rotate the file and try again, hoping it's transient
+ if (writeException.getMessage().contains("Stream Closed")) {
+ LOG.warn("Output Stream was closed. Attempting to rotate file and continue");
+ rotateOutputFile();
+ // If this write fails, the exception will be allowed to bubble up.
+ out.write(bytes);
--- End diff --
If there is a non-recoverable error, you are in the same position as before.
I was thinking of having a 'failed' flag ( similar to what hadoop has in it's screams where they track closed explicitly )
---
[GitHub] metron issue #741: METRON-1153 HDFS HdfsWriter never recovers from exception...
Posted by justinleet <gi...@git.apache.org>.
Github user justinleet commented on the issue:
https://github.com/apache/metron/pull/741
@ottobackwards From the ticket description, it looks like you thought there was a more involved refactoring for doing something like this. Are you good with this sort of solution (at least as an initial step)?
---
[GitHub] metron issue #741: METRON-1153 HDFS HdfsWriter never recovers from exception...
Posted by justinleet <gi...@git.apache.org>.
Github user justinleet commented on the issue:
https://github.com/apache/metron/pull/741
Adjusted to only catch the "Stream Closed" exception, not the general IOException. If it's stream closed, attempt the rotate and write, otherwise rethrow
---