[Console] Fix autocomplete_entities API crash when response size is too big #140569
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
e5ae3dd to
db5e2b5
Compare
alexwizp
reviewed
Sep 19, 2022
src/plugins/console/server/routes/api/console/autocomplete_entities/register_get_route.ts
Outdated
Show resolved
Hide resolved
alexwizp
reviewed
Sep 19, 2022
db5e2b5 to
01ae4c0
Compare
08a9cb0 to
434ac82
Compare
sabarasaba
reviewed
Oct 5, 2022
alexwizp
reviewed
Oct 5, 2022
alexwizp
reviewed
Oct 5, 2022
c4f19ef to
db281e3
Compare
alexwizp
reviewed
Oct 7, 2022
alexwizp
reviewed
Oct 7, 2022
491bf97 to
3b75822
Compare
|
Closing this PR as the issue was fixed by #143428 |
added 9 commits
November 10, 2022 10:37
2de7634 to
f94b90b
Compare
Thanks for the review @alisonelizabeth! I've addressed your feedback with f94b90b and a31e406. We have existing API integration tests for autocomplete_entities, but they are not very extensive. I've added a few more tests to cover the response format. |
|
Verified the api integration tests through the flaky test runner ~200 times https://buildkite.com/elastic/kibana-flaky-test-suite-runner/builds/1539 |
alisonelizabeth
approved these changes
Nov 10, 2022
💚 Build Succeeded
Metrics [docs]Async chunks
Page load bundle
Unknown metric groupsESLint disabled in files
ESLint disabled line counts
Total ESLint disabled count
History
To update your PR or re-run it, just comment with: cc @mibragimov |
kibanamachine
pushed a commit
to kibanamachine/kibana
that referenced
this pull request
Nov 11, 2022
…oo big (elastic#140569) Fixes elastic#144310 ### Summary This PR addresses the issue of the Kibana instance restarting when the response size is too big for the `autocomplete_entities` API. This happens when a cluster has a large number of mappings and we try to retrieve them all on the server side with `esClient.asInternalUser.indices.getMapping()`. esClient does not handle large responses well and throws an error that causes the Kibana instance to restart. As node's max [string length](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/String/length#description) is 2^28-1 (~512MB) if the response size is over 512MB, it will throw an error with the message "Invalid string length". The fix is to use the raw http request to retrieve the mappings instead of esClient and check the response size before sending it to the client. If the response size is too big, we will return an empty object and log the error in the server logs. #### Proposed changes - Remove ES JS client requests and use native Node.js HTTP client instead - Limit the response size to 10MB for the `autocomplete_entities` API #### How to test this PR locally To test this out, you will need to connect Kibana to a remote cluster with a large number of mappings. We created a patch file that you can apply to your local Kibana instance to test this PR. Since the patch file contains credentials, we can't share it publicly. Please reach out to me if you would like to test this PR. I will share the patch file and the instructions to apply it. Co-authored-by: Muhammad Ibragimov <muhammad.ibragimov@elastic.co> (cherry picked from commit 0e01401)
💚 All backports created successfully
Note: Successful backport PRs will be merged automatically after passing CI. Questions ?Please refer to the Backport tool documentation |
kibanamachine
added a commit
that referenced
this pull request
Nov 11, 2022
…e is too big (#140569) (#145051) # Backport This will backport the following commits from `main` to `8.5`: - [[Console] Fix autocomplete_entities API crash when response size is too big (#140569)](#140569) <!--- Backport version: 8.9.7 --> ### Questions ? Please refer to the [Backport tool documentation](https://github.com/sqren/backport) <!--BACKPORT [{"author":{"name":"Muhammad Ibragimov","email":"53621505+mibragimov@users.noreply.github.com"},"sourceCommit":{"committedDate":"2022-11-11T10:43:52Z","message":"[Console] Fix autocomplete_entities API crash when response size is too big (#140569)\n\nFixes https://github.com/elastic/kibana/issues/144310\r\n\r\n### Summary\r\n\r\nThis PR addresses the issue of the Kibana instance restarting when the\r\nresponse size is too big for the `autocomplete_entities` API. This\r\nhappens when a cluster has a large number of mappings and we try to\r\nretrieve them all on the server side with\r\n`esClient.asInternalUser.indices.getMapping()`. esClient does not handle\r\nlarge responses well and throws an error that causes the Kibana instance\r\nto restart. As node's max [string\r\nlength](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/String/length#description)\r\nis 2^28-1 (~512MB) if the response size is over 512MB, it will throw an\r\nerror with the message \"Invalid string length\".\r\n\r\nThe fix is to use the raw http request to retrieve the mappings instead\r\nof esClient and check the response size before sending it to the client.\r\nIf the response size is too big, we will return an empty object and log\r\nthe error in the server logs.\r\n\r\n#### Proposed changes\r\n\r\n- Remove ES JS client requests and use native Node.js HTTP client\r\ninstead\r\n- Limit the response size to 10MB for the `autocomplete_entities` API\r\n\r\n#### How to test this PR locally\r\nTo test this out, you will need to connect Kibana to a remote cluster\r\nwith a large number of mappings. We created a patch file that you can\r\napply to your local Kibana instance to test this PR. Since the patch\r\nfile contains credentials, we can't share it publicly. Please reach out\r\nto me if you would like to test this PR. I will share the patch file and\r\nthe instructions to apply it.\r\n\r\nCo-authored-by: Muhammad Ibragimov <muhammad.ibragimov@elastic.co>","sha":"0e0140181faf65ad3041e22fc053e49c76f2ce47","branchLabelMapping":{"^v8.6.0$":"main","^v(\\d+).(\\d+).\\d+$":"$1.$2"}},"sourcePullRequest":{"labels":["Feature:Console","Feature:Dev Tools","release_note:fix","Team:Deployment Management","backport:prev-minor","v8.6.0"],"number":140569,"url":"https://github.com/elastic/kibana/pull/140569","mergeCommit":{"message":"[Console] Fix autocomplete_entities API crash when response size is too big (#140569)\n\nFixes https://github.com/elastic/kibana/issues/144310\r\n\r\n### Summary\r\n\r\nThis PR addresses the issue of the Kibana instance restarting when the\r\nresponse size is too big for the `autocomplete_entities` API. This\r\nhappens when a cluster has a large number of mappings and we try to\r\nretrieve them all on the server side with\r\n`esClient.asInternalUser.indices.getMapping()`. esClient does not handle\r\nlarge responses well and throws an error that causes the Kibana instance\r\nto restart. As node's max [string\r\nlength](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/String/length#description)\r\nis 2^28-1 (~512MB) if the response size is over 512MB, it will throw an\r\nerror with the message \"Invalid string length\".\r\n\r\nThe fix is to use the raw http request to retrieve the mappings instead\r\nof esClient and check the response size before sending it to the client.\r\nIf the response size is too big, we will return an empty object and log\r\nthe error in the server logs.\r\n\r\n#### Proposed changes\r\n\r\n- Remove ES JS client requests and use native Node.js HTTP client\r\ninstead\r\n- Limit the response size to 10MB for the `autocomplete_entities` API\r\n\r\n#### How to test this PR locally\r\nTo test this out, you will need to connect Kibana to a remote cluster\r\nwith a large number of mappings. We created a patch file that you can\r\napply to your local Kibana instance to test this PR. Since the patch\r\nfile contains credentials, we can't share it publicly. Please reach out\r\nto me if you would like to test this PR. I will share the patch file and\r\nthe instructions to apply it.\r\n\r\nCo-authored-by: Muhammad Ibragimov <muhammad.ibragimov@elastic.co>","sha":"0e0140181faf65ad3041e22fc053e49c76f2ce47"}},"sourceBranch":"main","suggestedTargetBranches":[],"targetPullRequestStates":[{"branch":"main","label":"v8.6.0","labelRegex":"^v8.6.0$","isSourceBranch":true,"state":"MERGED","url":"https://github.com/elastic/kibana/pull/140569","number":140569,"mergeCommit":{"message":"[Console] Fix autocomplete_entities API crash when response size is too big (#140569)\n\nFixes https://github.com/elastic/kibana/issues/144310\r\n\r\n### Summary\r\n\r\nThis PR addresses the issue of the Kibana instance restarting when the\r\nresponse size is too big for the `autocomplete_entities` API. This\r\nhappens when a cluster has a large number of mappings and we try to\r\nretrieve them all on the server side with\r\n`esClient.asInternalUser.indices.getMapping()`. esClient does not handle\r\nlarge responses well and throws an error that causes the Kibana instance\r\nto restart. As node's max [string\r\nlength](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/String/length#description)\r\nis 2^28-1 (~512MB) if the response size is over 512MB, it will throw an\r\nerror with the message \"Invalid string length\".\r\n\r\nThe fix is to use the raw http request to retrieve the mappings instead\r\nof esClient and check the response size before sending it to the client.\r\nIf the response size is too big, we will return an empty object and log\r\nthe error in the server logs.\r\n\r\n#### Proposed changes\r\n\r\n- Remove ES JS client requests and use native Node.js HTTP client\r\ninstead\r\n- Limit the response size to 10MB for the `autocomplete_entities` API\r\n\r\n#### How to test this PR locally\r\nTo test this out, you will need to connect Kibana to a remote cluster\r\nwith a large number of mappings. We created a patch file that you can\r\napply to your local Kibana instance to test this PR. Since the patch\r\nfile contains credentials, we can't share it publicly. Please reach out\r\nto me if you would like to test this PR. I will share the patch file and\r\nthe instructions to apply it.\r\n\r\nCo-authored-by: Muhammad Ibragimov <muhammad.ibragimov@elastic.co>","sha":"0e0140181faf65ad3041e22fc053e49c76f2ce47"}}]}] BACKPORT--> Co-authored-by: Muhammad Ibragimov <53621505+mibragimov@users.noreply.github.com>
|
Changing |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #144310
Summary
This PR addresses the issue of the Kibana instance restarting when the response size is too big for the
autocomplete_entitiesAPI. This happens when a cluster has a large number of mappings and we try to retrieve them all on the server side withesClient.asInternalUser.indices.getMapping(). esClient does not handle large responses well and throws an error that causes the Kibana instance to restart. As node's max string length is 2^28-1 (~512MB) if the response size is over 512MB, it will throw an error with the message "Invalid string length".The fix is to use the raw http request to retrieve the mappings instead of esClient and check the response size before sending it to the client. If the response size is too big, we will return an empty object and log the error in the server logs.
Proposed changes
autocomplete_entitiesAPIHow to test this PR locally
To test this out, you will need to connect Kibana to a remote cluster with a large number of mappings. We created a patch file that you can apply to your local Kibana instance to test this PR. Since the patch file contains credentials, we can't share it publicly. Please reach out to me if you would like to test this PR. I will share the patch file and the instructions to apply it.