I have a lambda function (python) where I update existing dynamodb items using PartiQL. The update is executed in batches of 5, and I use the standard retry mode with 5 attempts:
from botocore.config import Config
config = Config(
retries = {
'max_attempts': 5,
'mode': 'standard'
}
)
In addition, I add a random sleep time of 20ms to 60ms between batches. Lambda is executed with a maximal number of 200 records each time.
I see that despite the retry mode and the random sleep time, I sometimes get this error:
The level of configured provisioned throughput for the table was exceeded. Consider increasing your provisioning level with the UpdateTable API.
The reason is not clear, because when I look at the monitoring tab of the dynamodb table, I can see that the write usage is well below the provisioning (image attached).
What can I do to solve it? Thanks.
I have a lambda function (python) where I update existing dynamodb items using PartiQL. The update is executed in batches of 5, and I use the standard retry mode with 5 attempts:
from botocore.config import Config
config = Config(
retries = {
'max_attempts': 5,
'mode': 'standard'
}
)
In addition, I add a random sleep time of 20ms to 60ms between batches. Lambda is executed with a maximal number of 200 records each time.
I see that despite the retry mode and the random sleep time, I sometimes get this error:
The level of configured provisioned throughput for the table was exceeded. Consider increasing your provisioning level with the UpdateTable API.
The reason is not clear, because when I look at the monitoring tab of the dynamodb table, I can see that the write usage is well below the provisioning (image attached).
What can I do to solve it? Thanks.
Share Improve this question edited Mar 13 at 9:41 DarkBee 15.5k8 gold badges72 silver badges118 bronze badges asked Mar 13 at 9:10 Mister_LMister_L 2,6119 gold badges37 silver badges68 bronze badges 2- 1 Seems like the graph shows averaage consumed capacity, not maximum. – gshpychka Commented Mar 14 at 14:24
- Do you get the same error if you change the billing type of the dynomoDB table to non-provisioned? – shantanuo Commented Mar 18 at 7:11
1 Answer
Reset to default 0The error is clear, you exceed your provisioned capacity. The CloudWatch graph is misleading as it only shows average consumption for a 1 minute period, but DynamoDB counts on a per second basis. Such that if you made 5000 requests in 1-3 seconds but nothing for the remainder of the minute, you'd get a much lower data point on CloudWatch.
To overcome this throttling, you can either scale your min autoscaling to a higher value, or more effective, use on-demand mode which scales instantly.