All Products
Search
Document Center

Agent Sandbox:Limits

Last Updated:Oct 13, 2026

Agent Sandbox Flash Edition and Container Edition have limits in areas such as quotas, runtime resources, and concurrency. Check the limits in this topic before you create resources to prevent your workloads from being blocked.

Flash Edition

Quota limits

Before you create sandbox resources, take note of the following quota limits:

Limit item

Default limit

Description

Maximum number of sandbox instances

300 instances

The maximum number of sandbox instances in a single region.

Maximum template image size

10 GB

The maximum size of a compressed template image.

Sandbox runtime resource limits

The following limits apply to a single sandbox instance at runtime:

Limit item

Default limit

Description

Instance concurrency

200

The maximum number of concurrent requests that a single sandbox instance supports. This limit cannot be adjusted.

Temporary disk space

15 GB

The disk space of a single sandbox instance.

File descriptors

100000

The number of file descriptors that a single sandbox instance can use. This limit cannot be adjusted.

Total number of processes and threads

1024

The total number of processes and threads that a single sandbox instance can use. This limit cannot be adjusted.

Maximum runtime in the active state

86400s

The maximum duration for which a single sandbox instance can run in the active state at a time.

Payload size of a synchronous request body

32 MB

The payload size of the request body for a single synchronous invocation. This limit cannot be adjusted.

Payload size of an asynchronous request body

128 KB

The payload size of the request body for a single asynchronous invocation. This limit can be adjusted up to a maximum of 256 KB.

Bandwidth

1 Gbit/s to 5 Gbit/s

The bandwidth range of a sandbox instance. This limit cannot be adjusted.

Sandbox instance delivery rate

The delivery rate of sandbox instances is subject to the following limit:

Limit item

Default limit

Description

Instance delivery rate

300 instances per minute

The number of sandbox instances that can be delivered per minute. This limit can be adjusted.

Container Edition

Limits

Before you use Alibaba Cloud Container Compute Service (ACS), be aware of the following limits:

  • After you create an ACS cluster, the following changes are currently not supported:

    • Change the Virtual Private Cloud (VPC) of the cluster.

    • Change the container network plug-in.

    • Change the storage plug-in.

    • Migrate applications between namespaces.

  • The following traffic limits apply when you access the control plane components of a cluster:

    Because of bandwidth limits, reading a large number of cluster events in a single API or command-line interface (CLI) request can throttle requests to the control plane components (API server and etcd). This can cause the read operation to fail. To prevent this, query cluster events in the ACS console by navigating to Operations > Event Center. Alternatively, add a pagination parameter, such as --chunk-size=500, to your API call or command to reduce the amount of data in a single request.

Quotas

Item

Upper limit

Default quota

How to increase

Total number of ACS clusters per account

20

3

You can go to Quota Center to apply for an increase to the default quota. Select Total Number of ACS Clusters as the quota name.

Note

If the upper limit cannot meet your business requirements, you can to contact technical support. For Product, select Container Service for Kubernetes.

Total number of pay-as-you-go elastic vCores per region

30,000 vCPUs

300 vCPUs

Go to Quota Center, select the target region, and request an increase to the default quota.

Note

If the upper limit cannot meet your business requirements, you can to contact technical support. For Product, select Alibaba Cloud Container Compute Service (ACS).

Capacity limits

Item

Upper limit

How to increase

etcd storage capacity

8 GB

Cannot be increased

Total size of each etcd object type

800 MB

Cannot be increased

Pods

50,000

Once this limit is reached, the system rejects requests to create new pods. You can to request a quota increase.

Important

If many pods are associated with services, keep the number below 20,000.

Services

10,000

Once this limit is reached, the system rejects requests to create new services. You can to request a quota increase.

Note

If a single namespace contains many services, the large volume of environment variables injected by kubelet can cause pods to start slowly or fail. To resolve this performance bottleneck, you can disable this injection by setting enableServiceLinks: false in the pod specification.

ConfigMaps

30,000

You can to request a quota increase.

Secrets

100,000

For ACS clusters that were created with a version earlier than 1.33 or were upgraded from a version earlier than 1.33, the quota for Secrets, PVCs, and PVs is 10,000. For these clusters, you can submit a ticket request technical support to increase the quota. However, the quota cannot be increased if it has already reached 100,000.

Persistent volume claims (PVCs)

100,000

Persistent volumes (PVs)

100,000

Total number of custom resources (CRs) of a single type

100,000

The default limit is 10,000 and can be raised to a maximum of 100,000 on request. Once this limit is reached, the system rejects requests to create new resources. You can to request a quota increase.

Total number of custom resource definitions (CRDs)

100,000

Cannot be increased

Concurrency limits

Image cache

The following table lists the maximum number of calls per minute for each API in a single region.

API

Max calls per minute

CreateImageCache

300

DeleteImageCache

300

ListImageCaches

600

GetImageCache

6,000

Instance delivery throughput

The following table describes the upper limits of instance-related throughput per minute for a single Alibaba Cloud account (primary account) in a single region (including internal retries). By default, all clusters in a region share the same limits. To request per-cluster quota allocation, you can submit a ticket to contact technical support.

ACS uses a token bucket algorithm per Alibaba Cloud account per region to throttle instance-related operations such as creating (starting), deleting, hibernating, and resuming container instances or pods. The token bucket capacity corresponds to the burst rate, which is the maximum number of requests allowed in a short period. The token fill rate corresponds to the continuous rate, which is the average request processing capacity over a long period.

Instance-related capability

Token bucket capacity (burst rate)

Token fill rate (continuous rate)

Instance creation

100

20

Instance deletion

100

20

Instance hibernation

50

10

Instance resumption

50

10

Checkpoint creation

50

10

Checkpoint deletion

50

10

Commit creation

50

10

Commit deletion

50

10

Underlying service quotas

Category

Item

Default quota

How to increase

Virtual Private Cloud (VPC) limits

The maximum number of custom route entries that you can create in a route table (excluding dynamically propagated route entries).

200

Request a quota increase on the Quota Management page or in Quota Center.

The maximum number of dynamically propagated route entries in a route table.

200

The maximum number of vSwitches that you can create in a VPC.

150

The maximum number of VPCs that you can create in a region.

10

The maximum number of private network addresses that cloud resources in a VPC can use.

300,000

1. If an ECS instance has only one private IP address, it consumes only one network address.
2. If an ECS instance is attached to multiple elastic network interfaces (ENIs) or an ENI has multiple IP addresses, the number of network addresses it consumes is the sum of all IP addresses on its attached ENIs.

No

Number of private IP addresses that a VPC-type security group can contain for a single Alibaba Cloud account in a specific region

  • Basic security groups:

    6000

    Note
    • Capacity usage is the total number of private IP addresses on all associated ENIs (primary and secondary), including primary private IPv4, IPv6, secondary private IPv4, IPv4 prefixes, and IPv6 prefixes.

    • If you need more than 6,000 private IP addresses to communicate over the internal network, assign the ECS instances to multiple security groups and add rules that authorize mutual access.

    • View the maximum private IP address count for a basic security group in a VPC in Quota Center with quota ID q_vpc-normal-security-group-ip-count.

  • Enterprise security groups:

    65,536

    Capacity usage is based on the total number of ENIs (primary and secondary) associated with the security group.

No

Maximum number of elastic network interfaces (secondary network interfaces) that can be created by a single Alibaba Cloud account in a specific region

View in Quota Center.

Go to Quota Center to request a quota increase.

Pay-as-you-go EIPs per account (excludes subscription EIPs)

20

Go to Quota Center to request a quota increase.

Server Load Balancer (SLB) limits

Maximum number of CLB instances that can be created by each Alibaba Cloud account

30

You can request a quota increase by using one of the following methods:

Maximum number of backend servers that can be added to a CLB instance

200

You can request a quota increase by using one of the following methods:

Maximum number of listeners that can be added to a CLB instance

50

You can request a quota increase by using one of the following methods:

Cloud Disk limits

Pay-as-you-go Cloud Disk quantity quota per account across all regions

You can view the quota in Quota Center.

You can to request a quota increase.

Pay-as-you-go Cloud Disk capacity quota for data disks per account

This quota depends on your cloud server usage, region, and Cloud Disk type. You can view the quota in Quota Center. For more information, see View or increase EBS quotas.

You can to request a quota increase.