Filter on size.value still uses greater than:

namespace filter is working in dv EKS:

@creynolds, the issue with the code is that the namespace file is being treated as a floating-point number. We don’t allow exact comparisons with floating-point values because they can’t be reliably compared due to how they’re stored.

DV QA testing, files after copy and aggregate copy:

I'm pausing additional QA testing for now, as the EKS DV Snapshots DAG is not functioning as expected. These issues appear to be unrelated to the group_1 ticket.

The ticket will need to test the Kyoto article and see if the Japanese characters need to be in the Abstract.

The issue stems from an edge case in the Kyoto article.

Scenario: A client is downloading an SC project (containing multiple chunks) while our clean-up process deletes some or all of those files. How can we ensure this operation is atomic and safe?

Waiting on T389548 so we can QA test the updated Cap configuration in service-deploy

Similarly, the aggregate JSON file has the source filenames in the identifier fields. Should the group_1 aggregate JSON have the updated identifier names?

@prabhat, regarding the snapshot metadata in these JSON files (see example screenshots), should the chunks array and the identifier fields be updated to include the _group_1 suffix in the copied S3 files?

Our HTTP connections to KSQLDB use connection reuse. When a connection is closed to free up resources, any reused connections are also terminated, leading to attempts to write data after the stream has been closed. We're using Envoy in K8s that manages connection pools, which can close idle TCP connections, this could be contributing to the closed connections in our Push code.

Started a discussion about how we can do consistent logging across all packages/submodules. Choices: Use go.mod replace command with submodules, or publish log submodule as a GitHub public OSS package.

Looks like resources freed up better than before, see first chart goes back to baseline after queries finish. Previously the query count stayed high for prolonged periods of time

No action required

@JArguello-WMF I created a MR to implement the changes suggested by Prabhat, pasting text from Slack here:

Bulk ingestion Structured-Data has revert risk in both ECS and EKS dv environments.

Adding reverrisk secret to article bulk in EKS, using repos: eks-clusters, service-deploy

Picking it up, now that most MRs are in prd

The bulk Ingestion works with revert risk, but the ML API times out, with a failure rate of 66% in dv. I've asked Prabhat and Francisco to decide if this is an acceptable error rate and propose solutions.

HubSpot uses a standard OAuth token for authentication (https://developers.hubspot.com/docs/guides/api/overview). Since the token will be stored in AWS Secrets Manager, it will not be exposed in client-side JavaScript or Go code.

@FNavas-foundation We moved to block until developers get clarification on which language to use with Reference Needed (use it with the 10 recommended languages, or the 5 languages that have published F1 scores)