[
https://issues.apache.org/jira/browse/BEAM-10990?focusedWorklogId=658976&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-658976
]
ASF GitHub Bot logged work on BEAM-10990:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 01/Oct/21 14:28
Start Date: 01/Oct/21 14:28
Worklog Time Spent: 10m
Work Description: egalpin commented on pull request #15381:
URL: https://github.com/apache/beam/pull/15381#issuecomment-932279563
@echauchot It's highly possible that I have a fundamental misunderstanding
about how to handle window data. I'm going to try to outline my goals and
challenges, and take the proposed implementation out of focus.
Goals:
- Change BulkIO/Write to output PCollectionTuple rather than PDone, in order
to support reporting the status of indexing each input document
- Leave windows/timestampes/etc of input data entirely unaltered
Challenges:
- BulkIOBaseFn relies on buffering inputs, either using bundles or Stateful
specs
- BulkIOBaseFn#finishBundle() must be called to ensure that any buffered
inputs are sent to ES, and as of this PR, output to the PCollectionTuple
- DoFn.FinishBundle mehtods can accept a
[FinishBundleContext](https://beam.apache.org/releases/javadoc/2.32.0/org/apache/beam/sdk/transforms/DoFn.FinishBundleContext.html)
in order to output elements
- FinishBundleContext output methods all require explicit specification of a
BoundedWindow instance
I got a bit stuck on that last point. My impression was that in order to
ensure buffered docs' results were output, and in order to leave those
elements' windows unaltered, I needed to keep track of the windows to which
those elements belong so that they could be explicitly passed to
FinishBundleContext#output.
I'd definitely be keen to learn more about how to handle windows and
challenge my assumptions here. Thanks for your time @echauchot in reviewing
and teaching.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 658976)
Time Spent: 4h 40m (was: 4.5h)
> Elasticsearch IO Infinite loop with write Error when the pipeline job
> streaming mode
> -------------------------------------------------------------------------------------
>
> Key: BEAM-10990
> URL: https://issues.apache.org/jira/browse/BEAM-10990
> Project: Beam
> Issue Type: Improvement
> Components: io-java-elasticsearch
> Affects Versions: 2.24.0
> Reporter: Steven Gaunt
> Assignee: Evan Galpin
> Priority: P3
> Time Spent: 4h 40m
> Remaining Estimate: 0h
>
> When streaming messages from PubsubIO , the pipeline is in Streaming mode.
> If for some reason the ElasticSearchIO.Write() has an response from the
> ElasticSearch index api, the writefn will throw a IOException. Since this
> excetpion is part of the Write transform, it becomes an unhandled error.
> This will then inheritly cause behaviour from the job pipeline to infinitely
> retry that error..
> _*snippet form beam website*_
> The Dataflow service retries failed tasks up to 4 times in batch mode, and an
> unlimited number of times in streaming mode. In batch mode, your job will
> fail; in streaming, it may stall indefinitely.
>
> This is the ElasticSearchIO.write transform .
> {code:java}
> public PDone expand(PCollection<String> input) {
> ElasticsearchIO.ConnectionConfiguration connectionConfiguration =
> this.getConnectionConfiguration();
> Preconditions.checkState(connectionConfiguration != null,
> "withConnectionConfiguration() is required");
> * input.apply(ParDo.of(new ElasticsearchIO.Write.WriteFn(this)));*
> return PDone.in(input.getPipeline());
> }
> {code}
> The pardo function (WriteFn) finishBundle step will call
> ElasticsearchIO.checkForErrors helper method which will throw exception if
> the http response from elasticsearch has error in the json reponse.
> {code:java}
> // Some comments here
> public void finishBundle(DoFn<String, Void>.FinishBundleContext context)
> throws IOException, InterruptedException {
> this.flushBatch();
> }
> private void flushBatch() throws IOException,
> InterruptedException {
> if (!this.batch.isEmpty()) {
> StringBuilder bulkRequest = new StringBuilder();
> Iterator var2 = this.batch.iterator();
> while(var2.hasNext()) {
> String json = (String)var2.next();
> bulkRequest.append(json);
> }
> this.batch.clear();
> this.currentBatchSizeBytes = 0L;
> String endPoint = String.format("/%s/%s/_bulk",
> this.spec.getConnectionConfiguration().getIndex(),
> this.spec.getConnectionConfiguration().getType());
> HttpEntity requestBody = new
> NStringEntity(bulkRequest.toString(), ContentType.APPLICATION_JSON);
> Request request = new Request("POST", endPoint);
> request.addParameters(Collections.emptyMap());
> request.setEntity(requestBody);
> Response response =
> this.restClient.performRequest(request);
> HttpEntity responseEntity = new
> BufferedHttpEntity(response.getEntity());
> if (this.spec.getRetryConfiguration() != null &&
> this.spec.getRetryConfiguration().getRetryPredicate().test(responseEntity)) {
> responseEntity = this.handleRetry("POST", endPoint,
> Collections.emptyMap(), requestBody);
> }
>
> ElasticsearchIO.checkForErrors((HttpEntity)responseEntity,
> this.backendVersion, this.spec.getUsePartialUpdate());
> }
> }
> static void checkForErrors(HttpEntity responseEntity, int backendVersion,
> boolean partialUpdate) throws IOException {
> JsonNode searchResult = parseResponse(responseEntity);
> boolean errors = searchResult.path("errors").asBoolean();
> if (errors) {
> StringBuilder errorMessages = new StringBuilder("Error writing to
> Elasticsearch, some elements could not be inserted:");
> JsonNode items = searchResult.path("items");
> Iterator var7 = items.iterator();
> while(var7.hasNext()) {
> JsonNode item = (JsonNode)var7.next();
> String errorRootName = "";
> if (partialUpdate) {
> errorRootName = "update";
> } else if (backendVersion == 2) {
> errorRootName = "create";
> } else if (backendVersion >= 5) {
> errorRootName = "index";
> }
> JsonNode errorRoot = item.path(errorRootName);
> JsonNode error = errorRoot.get("error");
> if (error != null) {
> String type = error.path("type").asText();
> String reason = error.path("reason").asText();
> String docId = errorRoot.path("_id").asText();
> errorMessages.append(String.format("%nDocument id %s: %s
> (%s)", docId, reason, type));
> JsonNode causedBy = error.get("caused_by");
> if (causedBy != null) {
> String cbReason = causedBy.path("reason").asText();
> String cbType = causedBy.path("type").asText();
> errorMessages.append(String.format("%nCaused by: %s
> (%s)", cbReason, cbType));
> }
> }
> }
> throw new IOException(errorMessages.toString());
> }
> }
> {code}
>
> As a possible suggestion, rather than throw the exception, could it be
> possible to write the exception to an errorhandling tupeltag which then can
> handled to a deadletter queue ?
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
