Relevant telegraf.conf
[agent]
interval = "10s"
round_interval = true
metric_batch_size = 1000
metric_buffer_limit = 10000
debug = true
logfile = "/tmp/telegraf/telegraf.log"
logformat = "structured"
#========================#
# INPUTS #
#========================#
[[inputs.cpu]]
percpu = true
totalcpu = true
collect_cpu_time = false
report_active = true
#========================#
# AGGREGATION #
#========================#
[[aggregators.starlark]]
period = "1m"
drop_original = true
script = "/etc/telegraf/twa.star"
#========================#
# OUTPUTS #
#========================#
[[outputs.file]]
files = ["/tmp/telegraf/metrics.out"]Logs from Telegraf
{"time":"2025-11-28T08:12:59.998229537Z","level":"DEBUG","msg":"Updated aggregation range [2025-11-28 08:12:00 +0000 UTC, 2025-11-28 08:13:00 +0000 UTC]"}
{"time":"2025-11-28T08:13:00.031147576Z","level":"DEBUG","msg":"MMM Metric(\"test_metric\", tags={}, fields={\"e\": 43}, time=1764317580031095589)"}
{"time":"2025-11-28T08:13:00.054546694Z","level":"DEBUG","msg":"Updated aggregation range [2025-11-28 08:13:00 +0000 UTC, 2025-11-28 08:14:00 +0000 UTC]"}
{"time":"2025-11-28T08:13:00.060504064Z","level":"DEBUG","msg":"MMM Metric(\"test_metric\", tags={}, fields={\"e\": 43}, time=1764317580060450475)"}
{"time":"2025-11-28T08:13:02.263268189Z","level":"DEBUG","msg":"Wrote batch of 2 metrics in 26.069838ms"}
System info
- Telegraf version: 1.36.4 - Platform: Docker Desktop on Windows - Docker image:
telegraf:latest
Docker
services:
telegraf:
image: telegraf:latest
container_name: telegraf-aggregator
volumes:
- ./telegraf.conf:/etc/telegraf/telegraf.conf:ro
- ./output_logs:/tmp/telegraf
- ./twa.star:/etc/telegraf/twa.star:ro
### Steps to reproduce
**twa.star**
```python
state = {}
def add(metric):
pass
def push():
metrics = []
m = Metric("test_metric")
m.fields["e"] = 43
print("MMM", m)
return m
def reset():
pass
-
Create output directory: mkdir output_logs
-
Run: docker-compose up
-
Wait for 2-3 minutes and observe the logs
-
Check the metrics output: cat output_logs/metrics.out
-
Check the debug logs: cat output_logs/telegraf.log | grep "MMM"
Notice the timestamps are duplicated (first two have same timestamp, next two have same timestamp, etc.).
Timeline breakdown:
08:12:59.998 - Aggregation range updated (still in previous period)08:13:00.031 - First push() execution → outputs first metric08:13:00.054 - Aggregation range updated (new period starts)08:13:00.060 - Second push() execution → outputs second metric08:13:02.263 - Batch of 2 metrics written (should be only 1!)
This pattern repeats at every minute boundary:
This pattern repeats at every minute boundary:
# Next period boundary
{"time":"2025-11-28T08:13:59.996166923Z","level":"DEBUG","msg":"Updated aggregation range [2025-11-28 08:13:00 +0000 UTC, 2025-11-28 08:14:00 +0000 UTC]"}
{"time":"2025-11-28T08:14:00.001036656Z","level":"DEBUG","msg":"MMM Metric(\"test_metric\", tags={}, fields={\"e\": 43}, time=1764317640001008349)"}
{"time":"2025-11-28T08:14:00.004767884Z","level":"DEBUG","msg":"Updated aggregation range [2025-11-28 08:14:00 +0000 UTC, 2025-11-28 08:15:00 +0000 UTC]"}
{"time":"2025-11-28T08:14:00.01071261Z","level":"DEBUG","msg":"MMM Metric(\"test_metric\", tags={}, fields={\"e\": 43}, time=1764317640010659408)"}
{"time":"2025-11-28T08:14:02.259031154Z","level":"DEBUG","msg":"Wrote batch of 2 metrics in 22.412102ms"}
Additional context
This issue appears to be specific to Docker Desktop on Windows. The double execution causes:
- Duplicate metrics being generated
- Incorrect aggregation results
- Resource waste from redundant processing
The timing suggests a potential race condition or improper synchronization around period boundary transitions in the aggregator plugin when running in this specific environment.
Expected behavior
Expected behavior
The aggregator's push() function should execute once per configured period (1 minute). Each period boundary should produce exactly one metric output.
Expected output in metrics.out
Should see one metric per period:
test_metric e=43i 1764317580000000000
test_metric e=43i 1764317640000000000
test_metric e=43i 1764317700000000000
Actual behavior
Actual behavior
The aggregator executes twice per period, producing duplicate metrics:
In metrics.out, you'll see duplicate entries:
test_metric e=43i 1764317580000000000
test_metric e=43i 1764317580000000000
test_metric e=43i 1764317640000000000
test_metric e=43i 1764317640000000000
test_metric e=43i 1764317643000000000
Two metrics per period (duplicates):
test_metric e=43i 1764317580000000000
test_metric e=43i 1764317580000000000
test_metric e=43i 1764317640000000000
test_metric e=43i 1764317640000000000
test_metric e=43i 1764317643000000000
Additional info
No response
Relevant telegraf.conf
Logs from Telegraf
System info
telegraf:latestDocker
services:
telegraf:
image: telegraf:latest
container_name: telegraf-aggregator
volumes:
- ./telegraf.conf:/etc/telegraf/telegraf.conf:ro
- ./output_logs:/tmp/telegraf
- ./twa.star:/etc/telegraf/twa.star:ro
Create output directory:
mkdir output_logsRun:
docker-compose upWait for 2-3 minutes and observe the logs
Check the metrics output:
cat output_logs/metrics.outCheck the debug logs:
cat output_logs/telegraf.log | grep "MMM"Notice the timestamps are duplicated (first two have same timestamp, next two have same timestamp, etc.).
Timeline breakdown:
08:12:59.998- Aggregation range updated (still in previous period)08:13:00.031- First push() execution → outputs first metric08:13:00.054- Aggregation range updated (new period starts)08:13:00.060- Second push() execution → outputs second metric08:13:02.263- Batch of 2 metrics written (should be only 1!)This pattern repeats at every minute boundary:
This pattern repeats at every minute boundary:
Additional context
This issue appears to be specific to Docker Desktop on Windows. The double execution causes:
The timing suggests a potential race condition or improper synchronization around period boundary transitions in the aggregator plugin when running in this specific environment.
Expected behavior
Expected behavior
The aggregator's
push()function should execute once per configured period (1 minute). Each period boundary should produce exactly one metric output.Expected output in metrics.out
Should see one metric per period:
Actual behavior
Actual behavior
The aggregator executes twice per period, producing duplicate metrics:
In metrics.out, you'll see duplicate entries:
Two metrics per period (duplicates):
Additional info
No response