Under prolonged load the switch stops routing. My questions are:
Is there a recognised fix for this?
If there is no fix, is this a recognised problem which other Netgear switches do not have?
If neither of the above are the case, I assume that means I need to look at other manufacturers products.
In more detail
Tested with JGS524PE - 24-Port Gigabit Ethernet PoE Smart Managed Plus Switch. I have two of these at separate sites. Both exhibit the same behaviour. Swapping the switches makes no difference. Both were purchased new 18 months ago.
Typically after switching fails: no device can communicate with any other device connected to the switch, but I can still login to the JGS524PE web interface, and System -> Monitoring -> Port Statistics -> Refresh shows ongoing traffic. On a couple of occasions only communication with the high traffic device (NAS) was lost. On a couple of early occasions all communication to the JGS524PE was lost however I can't remember if that was only before I set the JGS524PE to have a static rather than DHCP set IP address.
JGS524PE switch function is restored by: Web interface -> System -> Maintenance -> Device Reboot or cycling power on the JGS524PE. Restarting devices attached to the JGS524PE is less reliable.
I my set up the fault is typically revealed once a month when software running on Windows (Arcserve StorageCraft ImageManager) creates / consolidates monthly backups on the NAS (Synology DS218 with 2 x 16TB drives). It occurs at both sites but more commonly a the site with larger monthly backups. At one site Image manager runs on Window 7, at the other site on a Windows 8.1 virtual machine or Windows 2012 standard virtual machine (the virtual machines running on Windows 2012 standard).
I have tried:
JGS524PE -> System → Maintenance -> defaults settings
Upgrading the firmware to the current 2.6.0.48 version
LAG has not been used
JGS524PE -> System → Multicast → IGMP Snooping → disable all
JGS524PE -> System → Management → Broadcast Forwarding → Dynamic (default) changed to Hardware
JGS524PE -> System → Management → Management mode → Web only (default as “Plus Utility support” exposes security vulnerabilities)
We have not seen any reported issues about this. I would definitely suggest getting in touch with Support by creating a case via online or phone so they can gather more information and try to troubleshoot or replicate before submitting this issue.
Just to update in case others have a similar problem, I think I have found a work around.
In additions to
JGS524PE -> System → Maintenance -> defaults settings
Upgrading the firmware to the current 2.6.0.48 version
LAG has not been used
JGS524PE -> System → Multicast → IGMP Snooping → disable all
JGS524PE -> System → Management → Broadcast Forwarding → Dynamic (default) changed to Hardware
JGS524PE -> System → Management → Management mode → Web only (default as “Plus Utility support” exposes security vulnerabilities)
I disabled
JGS524PE -> System → Management → Loop detection -> disable
Which appears to have prevented to lock ups. Unfortunately I can not be certain as in the past the lock ups have only occurred about once a month, so it is relatively early days atm.
As for the reason loop detection would lock up the device.
The desk top computers are wired to the LAN network but do have wifi hardware. However there is no access point at that site so should not have carried any traffic and besides the desktops did not loose LAN access first. Wifi in the desktops has now been disabled as well.
The lock ups appear to start between the NAS and windows server. The windows server has dual NIC and both are connected but only 1 is configured to be used. So while a small amount of broadcast traffic may go out through the second NIC, it should be minimal. And even if this minimal traffic under heavy load is enough to legitimately block traffic between the windows server and NAS, it should spread to all other ports (including direct IP address access).
In my opinion that leaves a firmware race condition bug. It would be nice is future firmware indicated when a port was disabled due to loop detection or other blocks such as broadcast forwarding / IGMP snooping was dropping packets.
Hopefully the end result is a functional router even if disabling most of it's features is required to achieve reliable function.
Thanks for the suggestion to contact support. I did and they tried to help, suggesting replacing one of the devices. I'm not convinced that would have had a high chance of success as the fault occurs with 2 devices at two sites despite changing devices. But at least they were friendly and made an effort. ATM they are short on replacement hardware but I will revisit their options in the future if required.
The simple loop detection does send out a certain packet on each port. If this packet does return on another port, the Plus switches disable one of the ports affected.
Either disconnect the unconfigured windows server adapter from the switch, or disable the loop detection as you have already done. Chime back if the "issue" still exists.
