ui: overload dashboard fails to render on large clusters #101377
Assignees
Labels
A-observability-inf
C-bug
Code not up to spec/doc, specs & docs deemed correct. Solution expected to change code/behavior.
Comments
dhartunian
added
C-bug
craig bot
pushed a commit
that referenced
this issue
Apr 20, 2023
101937: catalog: utilize scans for a large number of descriptors r=fqazi a=fqazi Previously, the catalog descriptor would fetch descriptors via point lookups using Get when scanning large batches of descriptors. This was further extended to also look up ZoneConfigs and comments in a similar way. Recently, we we started seeing regression on the Django test suite involving the pg_catalog tables, which tend to do read large number of descriptors, likely linked to extra overhead linked to both comments and zone configs in 23.1. To address this, this patch ill now start using scans for runs of descriptor IDs for batch scans which reduces the overall cost of fetching a large number of descriptors hiding this cost. Fixes: #100871 Release note: None 101943: build,release: provide way to inject build tag override, use in MAPB r=rail a=rickystewart When we build nightly builds, we build them in "release" configuration. Because we do this, the build tag reported by these binaries from `cockroach version` has been identical to that of an actual release binary, which is very confusing. Here we update the script to build these nightlies (`make-and-publish-build-artifacts.sh`) to inject an appropriate, identifiable build tag. It is probably wrong that we are building these nightlies as if they were "releases". We'll follow up with work to fix this and refine the build tags further. Closes #100532. Epic: None Release note (build change): Update reported `Build Tag` for nightly (non-release) builds 101944: sem/tree: remove TODO r=mgartner a=mgartner This commit removes a TODO that was addressed by #96045. Epic: None Release note: None 101952: ui: fix linegraph render for large clusters r=zachlite a=zachlite We were feeding `Math.min` and `Math.max` very large arrays (length equal to ~10e7). These functions take variadic arguments, and the runtime can't handle that many arguments. As a result we blow the stack, causing uncaught range errors. Instead, we'll use lodash min and max which are not variadic. Resolves #101377 Epic: None Release note: None Reproduction and Fix demonstrated on the command line: <img width="722" alt="Screenshot 2023-04-20 at 4 56 41 PM" src="https://user-images.githubusercontent.com/5423191/233486016-fd740a98-9a0d-4fef-a6d8-67e9cb4e318e.png"> Co-authored-by: Faizan Qazi <[email protected]> Co-authored-by: Ricky Stewart <[email protected]> Co-authored-by: Marcus Gartner <[email protected]> Co-authored-by: Zach Lite <[email protected]>
blathers-crl bot
pushed a commit
that referenced
this issue
Apr 20, 2023
We were feeding `Math.min` and `Math.max` very large arrays (length equal to ~10e7). These functions take variadic arguments, and the runtime can't handle that many arguments. As a result we blow the stack, causing uncaught range errors. Instead, we'll use lodash min and max which are not variadic. Resolves #101377 Epic: None Release note: None
blathers-crl bot
pushed a commit
that referenced
this issue
Apr 20, 2023
We were feeding `Math.min` and `Math.max` very large arrays (length equal to ~10e7). These functions take variadic arguments, and the runtime can't handle that many arguments. As a result we blow the stack, causing uncaught range errors. Instead, we'll use lodash min and max which are not variadic. Resolves #101377 Epic: None Release note: None
blathers-crl bot
pushed a commit
that referenced
this issue
Apr 20, 2023
We were feeding `Math.min` and `Math.max` very large arrays (length equal to ~10e7). These functions take variadic arguments, and the runtime can't handle that many arguments. As a result we blow the stack, causing uncaught range errors. Instead, we'll use lodash min and max which are not variadic. Resolves #101377 Epic: None Release note: None
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
On a cluster with 99 nodes, the overload dashboard failed to render with a stack overflow error in the javascript console. It was unclear what specifically caused this to happen and steps need to be taken to attempt to reproduce this error on a demo or test cluster.
Jira issue: CRDB-26919
Epic: CRDB-25476
The text was updated successfully, but these errors were encountered: