On ACX7000 platforms, egress scheduling is governed by the Enhanced Traffic Manager (ETM), which uses strict priority scheduling for priority queues and weighted fair queuing (WFQ) only for low-priority traffic. When a high-priority queue is configured without a shaping-rate, it can consume available bandwidth and starve lower-priority queues and router-generated control traffic. This behavior can result in protocol instability and control-plane packet loss. The issue was observed in a configuration where the TOP-PRIORITY queue was configured as a high-priority queue without a shaping-rate.

The following symptoms may be observed:

• Link Fault Management (LFM) and other router-generated control traffic experience packet loss or instability.
• High-priority traffic consumes excessive egress bandwidth during congestion.
• Lower-priority queues receive insufficient scheduling opportunities and may become starved.
• Control packets may be delayed or dropped even though interface bandwidth appears available.
• QoS behavior differs from MX platforms because ACX7000 does not support the MX-style in-profile/out-of-profile DWRR protection model for priority queues. Instead, strict priority queues continue to preempt lower priorities unless explicitly limited using shaping-rate.

The TOP-PRIORITY scheduler was configured with priority high but no shaping-rate:

set class-of-service schedulers TOP-PRIORITY buffer-size percent 70

set class-of-service schedulers TOP-PRIORITY priority high

Because ACX7000 uses strict priority scheduling, a priority queue without shaping can continuously receive ETM credits and dominate egress bandwidth, negatively affecting lower-priority and control traffic

Configure shaping-rate on all high-priority queues (except where platform design dictates otherwise) to prevent bandwidth starvation and protect lower-priority traffic. Juniper validation and JTAC lab testing recommend:

1. Configure router control traffic in the low-latency (LLQ) queue.
2. Apply an appropriate shaping-rate to high-priority data queues.
3. Tune the shaping-rate based on packet size, encapsulation overhead, and deployment requirements.
4. For 100G interfaces, a shaping-rate between 80% and 95% is generally recommended for high-priority traffic, depending on packet size and encapsulation.

set class-of-service schedulers GENERAL transmit-rate percent 15

set class-of-service schedulers GENERAL buffer-size percent 15

set class-of-service schedulers GENERAL priority low

set class-of-service schedulers MED-PRIORITY transmit-rate percent 10

set class-of-service schedulers MED-PRIORITY buffer-size percent 10

set class-of-service schedulers MED-PRIORITY priority low

set class-of-service schedulers NETWORK-CONTROL shaping-rate percent 5

set class-of-service schedulers NETWORK-CONTROL buffer-size percent 5

set class-of-service schedulers NETWORK-CONTROL priority low-latency

set class-of-service schedulers TOP-PRIORITY shaping-rate percent 85

set class-of-service schedulers TOP-PRIORITY buffer-size percent 70

set class-of-service schedulers TOP-PRIORITY priority high

This configuration was validated in the JTAC lab and ensures that control traffic receives the highest scheduling priority while preventing high-priority data traffic from monopolizing egress bandwidth.

Always configure a shaping-rate on ACX7000 priority queues.

Unlike MX platforms, ACX7000 priority queues do not have an in-profile/out-of-profile protection mechanism. The shaping-rate is the primary method to limit priority queue bandwidth consumption and prevent starvation of lower-priority services.