topic Re: Found inconsistency of SXL instances between SGMs in Hyperscale Firewall (Maestro)

Found inconsistency of SXL instances between SGMs

kamilazat — Thu, 20 Feb 2025 11:59:08 GMT

Hi guys!

It looks like asg perf has found some inconsistencies in SLX instances. Here's the output:

Performance Summary +--------------------------------------------+---------------+ |Name |Value | +--------------------------------------------+---------------+ |Throughput |4.4 G | |Packet rate |603.3 K | |Connection rate |9.7 K | |Concurrent connections |825.5 K | |Load average |8% | |Acceleration load (avg/min/max) |11%/10%/15% | |Instances load (avg/min/max) |9%/3%/16% | |Memory usage |17% | +--------------------------------------------+---------------+ * Instances / Acceleration Cores: 44 / 16 (!) * ATTENTION: Found inconsistency of SXL instances between SGMs * Activated SWB: FW Per SGW Distribution Summary +-----+----------+---------+------+-----------+-----------+-----------+-------+ |SGW |Throughput|Packet |Conn. |Concurrent |Accel. |Instances |Mem. | |ID | |rate |rate |Conn. |Cores |Cores |usage %| | | | | | |usage % |usage % | | +-----+----------+---------+------+-----------+-----------+-----------+-------+ |1_01 |1.8 G |277.1 K |4.9 K |364.3 K |11/10/15 |9/4/18 |16 | |1_02 |2.6 G |326.2 K |4.8 K |461.2 K |12/11/15 |9/3/15 |18 | +-----+----------+---------+------+-----------+-----------+-----------+-------+ |Total|4.4 G |603.3 K |9.7 K |825.5 K |11/10/15 |9/3/16 |17 | +-----+----------+---------+------+-----------+-----------+-----------+-------+ * Core usage shows avg, min, or max, as sampled during last 10 seconds

We made sure that fwaffinity.conf file has "i default auto" on both SGMs (nobody has ever meddled with that setting anyway).

Where does it find those 'inconsistencies' in that setup?

Cheers!

Re: Found inconsistency of SXL instances between SGMs

emmap — Fri, 21 Feb 2025 03:00:18 GMT

If dynamic balancing is enabled (default in R81.20+) then this is expected behaviour.

Re: Found inconsistency of SXL instances between SGMs

kamilazat — Mon, 24 Feb 2025 06:49:44 GMT

So it finds discrepancies between SXL instances because the core affinities are dynamically changed depending on the characteristics of the traffic. Do I get this correctly?

Re: Found inconsistency of SXL instances between SGMs

emmap — Mon, 24 Feb 2025 10:35:00 GMT

Yes, that's correct.

Re: Found inconsistency of SXL instances between SGMs

yura_k — Mon, 10 Mar 2025 06:37:40 GMT

starting from the JHF98, this information disappeared from the output:

The information about cores in the “asg perf” output is only get from the local SGM. Accordingly, warnings about mismatches with other SGMs simply cannot occur

Re: Found inconsistency of SXL instances between SGMs

PeterTheGreat — Thu, 15 May 2025 13:49:46 GMT

Hello, Yura!

It is applicable only to mixed SG.