Rebuild index job failed

[victorlin]

First failed on yesterday's scheduled run:

inventory for core - manifest key: nextstrain-data/config-v1/2024-08-08T01-00Z/manifest.json
There was an error while fetching the S3 inventory for core. This is fatal.
The manifest file for the S3 inventory for bucket nextstrain-data includes more than one inventory file. This situation was not encountered during development, but this is presumably caused by the inventory size exceeding some threshold and being chunked into multiple files. Please check this is indeed the case and, if so, amend the code to parse and join each file.

I noticed when pushing ceb01dc.

Progress

• Temporarily bypass broken resource indexer #976
• Support multiple inventory files #979
• Restore resource indexer dependency in CI #981

[victorlin]

I need the canary build to complete in order to complete the work associated with the commit, so I'll investigate.

[victorlin]

Comparing the manifest used for the previous commit on master to the current one:

diff \
    <(aws s3 cp s3://nextstrain-inventories/nextstrain-data/config-v1/2024-07-30T01-00Z/manifest.json -) \
    <(aws s3 cp s3://nextstrain-inventories/nextstrain-data/config-v1/2024-08-08T01-00Z/manifest.json -)

Indeed there are two inventory files now:

5c5
<   "creationTimestamp" : "1722301200000",
---
>   "creationTimestamp" : "1723078800000",
9,11c9,15
<     "key" : "nextstrain-data/config-v1/data/ace7ffe4-a9de-4cae-a17e-af5102e61a82.csv.gz",
<     "size" : 25209320,
<     "MD5checksum" : "86f6b86c69c1080abc25dcce78c2f4ed"
---
>     "key" : "nextstrain-data/config-v1/data/027be65e-df06-4962-8233-bd87cf606c08.csv.gz",
>     "size" : 24492873,
>     "MD5checksum" : "159639032cb63deac7abeba189b2af51"
>   }, {
>     "key" : "nextstrain-data/config-v1/data/48f7d241-e89b-4765-8d04-8a446371d0a0.csv.gz",
>     "size" : 945430,
>     "MD5checksum" : "4b9c6e8cd352124ed608515aa2575f8b"

[victorlin]

I have a local in-progress draft but I think it will require some testing and reviews which will extend into next week.

Since this is irrelevant to the group update, I'm going to temporarily bypass the index-resources job to unblock deployment.

[tsibley]

Previous context.