feat(users/flokli/nixos/archeology-ec2): automate bucket log parsing

This adds a `parse-bucket-logs.{service,timer}`, running once every
night at 3AM UTC, figuring out the last time it was run and parsing
bucket logs for all previous days.

It invokes the `archeology-parse-bucket-logs` script to produce
a .parquet file with the bucket logs in `s3://nix-cache-log/log/` for
that day (inside a temporary directory), then on success uploads the
produced parquet file to
`s3://nix-archeologist/nix-cache-bucket-logs/yyyy-mm-dd.parquet`.

Change-Id: Ia75ca8c43f8074fbaa34537ffdba68350c504e52
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10011
Reviewed-by: edef <edef@edef.eu>
Tested-by: BuildkiteCI
This commit is contained in:
Florian Klink 2023-11-12 18:09:03 +02:00 committed by flokli
parent 3fe455cd4a
commit e4adca0880
2 changed files with 80 additions and 0 deletions

View file

@ -6,6 +6,24 @@
../profiles/archeology.nix
];
systemd.timers.parse-bucket-logs = {
wantedBy = [ "multi-user.target" ];
timerConfig.OnCalendar = "*-*-* 03:00:00 UTC";
};
systemd.services.parse-bucket-logs = {
path = [ depot.users.flokli.archeology.parse-bucket-logs ];
serviceConfig = {
Type = "oneshot";
ExecStart = (pkgs.writers.writePython3 "parse-bucket-logs-continuously"
{
libraries = [ pkgs.python3Packages.boto3 ];
} ./parse-bucket-logs-continuously.py);
DynamicUser = "yes";
StateDirectory = "parse-bucket-logs";
};
};
environment.systemPackages = [
depot.users.flokli.archeology.parse-bucket-logs
];

View file

@ -0,0 +1,62 @@
import boto3
import datetime
import os
import re
import subprocess
import tempfile
s3 = boto3.resource('s3')
bucket_name = "nix-archeologist"
prefix = "nix-cache-bucket-logs/"
bucket = s3.Bucket(bucket_name)
key_pattern = re.compile(r'.*\/(?P<y>\d{4})-(?P<m>\d{2})-(?P<d>\d{2})\.parquet$') # noqa: E501
# get a listing (which is sorted), grab the most recent key
last_elem = list(
o for o in bucket.objects.filter(Prefix=prefix)
if key_pattern.match(o.key)
).pop()
# extract the date of that key.
m = key_pattern.search(last_elem.key)
last_elem_date = datetime.date(int(m.group("y")), int(m.group("m")), int(m.group("d"))) # noqa: E501
# get the current date (UTC)
now = datetime.datetime.now(tz=datetime.UTC)
now_date = datetime.date(now.year, now.month, now.day)
while True:
# Calculate what date would be processed next.
next_elem_date = last_elem_date + datetime.timedelta(days=1)
# If that's today, we don't want to process it.
if next_elem_date == now_date:
print("Caught up, would process data from today.")
break
# If we'd be processing data from yesterday, but it's right after midnight,
# also don't process - data might still be flushed.
if (next_elem_date + datetime.timedelta(days=1) == now_date) and now.hour == 0: # noqa: E501
print("Not processing data from previous day right after midnight")
break
src = f"http://nix-cache-log.s3.amazonaws.com/log/{next_elem_date.isoformat()}-*" # noqa: E501
# Invoke parse-bucket-logs script inside a tempdir and upload on success.
with tempfile.TemporaryDirectory() as td:
work_file_name = os.path.join(td, "output.parquet")
args = ["archeology-parse-bucket-logs", src, work_file_name]
subprocess.run(
args,
check=True # throw exception if nonzero exit code
)
dest_key = f"{prefix}{next_elem_date.isoformat()}.parquet"
# Upload the file
print(f"uploading to s3://{bucket_name}{dest_key}")
bucket.upload_file(work_file_name, dest_key)
last_elem_date = next_elem_date