API requests to endpoint "/api/v2/backup/metadata" result in file handle resource leakage

[cbainlive]

Steps to reproduce:

1. Run command sudo lsof -p `pidof influxd` | grep deleted | wc -l and note the count of deleted file handles held open by influxd
2. Run curl "http://localhost:8086/api/v2/backup/metadata" -XGET --header "Authorization: Token $TOKEN" --output response to hit the API endpoint
3. Re-run command sudo lsof -p `pidof influxd` | grep deleted | wc -l and note count has increased by 1

For example:

influxdb@localhost> uname -srm
Linux 5.4.0-148-generic x86_64
influxdb@localhost> influxd version
InfluxDB v2.6.1 (git: 9dcf880fe0) build_date: 2022-12-29T15:53:07Z
influxdb@localhost> sudo lsof -p `pidof influxd` | grep deleted | wc -l
0
influxdb@localhost> curl "http://localhost:8086/api/v2/backup/metadata" -XGET --header "Authorization: Token $TOKEN" --output response
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  278k    0  278k    0     0  22.6M      0 --:--:-- --:--:-- --:--:-- 22.6M
influxdb@localhost> sudo lsof -p `pidof influxd` | grep deleted | wc -l
1

Expected behavior:
The influxd process should clean up resources after completing processing of the API request.

Actual behavior:
A deleted file handle is left open each time the API endpoint is hit.

Environment info:

• System info: Linux 5.4.0-148-generic x86_64
• InfluxDB version: InfluxDB v2.6.1 (git: 9dcf880fe0) build_date: 2022-12-29T15:53:07Z

Config:
Default

[cbainlive]

In addition to the file descriptor leaking, there is disk space consumed by the OS as long as the process has it open.

I looked at the code and noticed that the issue is the File.Close method is not getting called correctly for this (and two other places) as implemented with a deferred call wrapping it with errors2.Capture. I have submitted a PR for it. Please consider adding this into the next 2.x build. Thanks.

[cbainlive]

Hi, pinging on this ticket. I provided the fix and I'm hoping this can make it into the next 2.x build since I'm having trouble building myself from code--I am also hitting this issue: #24363.