Scheduled configuration is ignored and app scales down #217
Description
We have the following configuration, which contains a recurring schedule:
{
"instance_min_count": 28,
"instance_max_count": 36,
"scaling_rules": [
{
"metric_type": "throughput",
"breach_duration_secs": 300,
"threshold": 60,
"operator": "<",
"cool_down_secs": 300,
"adjustment": "-1"
},
{
"metric_type": "throughput",
"threshold": 150,
"operator": ">=",
"cool_down_secs": 300,
"adjustment": "+12"
},
{
"metric_type": "throughput",
"breach_duration_secs": 60,
"threshold": 90,
"operator": ">=",
"cool_down_secs": 60,
"adjustment": "+1"
}
],
"schedules": {
"timezone": "Europe/Amsterdam",
"recurring_schedule": [
{
"start_time": "00:00",
"end_time": "23:59",
"days_of_month": [
19,
20,
21,
22,
23,
24,
25,
26,
27
],
"instance_min_count": 36,
"instance_max_count": 36
}
],
"specific_date": [
{
"start_date_time": "2025-11-28T00:00",
"end_date_time": "2025-12-04T23:59",
"instance_min_count": 36,
"instance_max_count": 36
}
]
}
}I expect that my application instances stay at 36 during the recurring schedule, but it happens that our application does start to scale down to the "default" minimum of 28 if the TPS is low enough.
For some unknown reason this does not always happen. We run two foundations and this has happened on one, another time on the other, and sometimes on both foundations.
| Event | Trigger | Date | Error |
|---|---|---|---|
| Instances scaled down from 29 to 28 | -1 instance(s) because throughput < 60rps for 300 seconds | Oct 22, 2025, 12:17:28 PM | |
| Instances scaled down from 30 to 29 | -1 instance(s) because throughput < 60rps for 300 seconds | Oct 22, 2025, 12:11:28 PM | |
| Instances scaled down from 31 to 30 | -1 instance(s) because throughput < 60rps for 300 seconds | Oct 22, 2025, 12:05:28 PM | |
| Instances scaled down from 32 to 31 | -1 instance(s) because throughput < 60rps for 300 seconds | Oct 22, 2025, 11:59:28 AM | |
| Instances scaled down from 33 to 32 | -1 instance(s) because throughput < 60rps for 300 seconds | Oct 22, 2025, 11:53:28 AM |
Note: the next night at 0:00 the recurring schedule is resumed, so the issue is always temporary.