Switch not pinging connected device?

Question

Computer networks

So.
There is an AT-8000GS/24POE switch (Allied Telesis). 24 copper and 4 optical ports.
The ports are in different VLANs. The switch also has its own subnet/VLAN.
Problem - communication with devices connected to ports under the same VLAN was lost.
If you pull out the patch cord from the switch and connect it to a laptop with an address from the lost subnet, then the device is seen, pinged, you can go to the web interface.
If you connect a laptop to any of the ports united by a common VLAN, then there is a connection with the device.
The port indication on the switch also indicates the working links on the ports.

POE indication

True, the POE indicator speaks of

An overload, short circuit,
or external voltage has been detected on the port

forum.alliedtelesis.ru/MY/Translation/Install_Guid... - 23rd page.
But it also flashes every 3 seconds. However, devices on POE ports do not consume and work more than a hundred in similar cases. Turned off power to the port.

However, if you go to the switch via SSH and give the PING command to the IP of the problem VLAN, then there is no connection, timeout and the address is unavailable.
Ports in access mode. Links are confirmed by both software and hardware. ACL is not present, addresses are static. Ports are marked as untagged. I don't even know where to look anymore.

We complicate

Коммутатор находится в стеке с 5-ю аналогичными. Первый, вышеописанный, соединён оптикой с главным оптическим коммутатором.
Оптический порт - trunk, all vlan, в качестве шлюза адрес оптического коммутатора из подсети коммутаторов, который является маршрутизатором (прописаны все VLAN и ассоциируемые с ними IP-адреса). К оптическому подсоединён третий коммутатор AT-8000GS/24POE, а к нему сервер (порт access, проблемный VLAN). Сервер видит устройства на других коммутаторах этого VLAN, пингует их и общий шлюз на оптическом коммутаторе. Виден также проблемный коммутатор. Брандмауэр вырублен, антивируса нет.

From the ports of the problem switch, the server from the same subnet is not visible. If you register a problem vlan free port of another switch in the stack, then the server pings, but the connected device is not visible from the server.

UPD 1 I drew approximately a diagram:

All switches see each other, except for those that do not ping. These are 2 intermediate switches, ping comes from them, but not to them.
From 10.5.0.110, ping goes all the way to the ACS server, in both directions, and from the switch itself only to the gateway 10.0.128.1, and the laptop in the 7th VLAN up to the server.
What I did: deleted / re-registered VLAN 7 on 10.5.0.110 - did not help.

UPD 2Moved the problematic devices to the next stack. Registered VLAN 7. Earned. The connection scheme is no different, the settings are the same. Only optical ports are nearby. I will understand and diagnose ports.

UPD3
Probably found a problem, but I can not solve it yet. AT-x610-24SPs/X IP Gateway 10.5.0.1 Layer 3 Refine
Both intermediate switches 10.5.0.92 and 10.5.0.97 have 10.5.0.0 as the gateway. ARP tables of both are empty.
forum.alliedtelesis.ru/viewtopic.php?f=6&t=3349 Asked a question on the manufacturer's forum.

Destination Mask NextHop Interface Protocol

---------------------------------------------------- -----------------------

10.5.0.0 255.255.255.0 10.5.0.92 vlan5-0 INTERFACE

UPD4 The
ball began to unravel. The value became the key: Destination 10.5.0.0
10.5.0.0 - specified as a gateway. I began to understand - and maybe there is a device with such an address on the network - it turned out not.
10.5.0.0 is a reserved network address 10.5.0.1-10.5.0.254
So the gateway is not specified at all.
This tells me personally that routing on a cool managed switch works like a dumb hub, well, or semi-dumb.
Next : switches 10.5.0.92 and 10.5.0.97 are optical, with a total of 4 redundant (duplicate) copper ports. You just don't connect with them. They are not on the web.
Connected via SSH using a USB converter to the console port. He pulled out one of the trunk optical ports. Altered the trunk copper port duplicating the access mode. I connected with a laptop with the set address 10.5.0.15 255.255.255.0 - no connection. Registered with the gateway 10.5.0.92 (experimented on it) and VOILA!

There is no ping with the switch, but there were several similar devices that were not visible on the 10.5.0.0 network.
Then I managed to compare the config 10.5.0.92 with a completely similar switch. Found a lot of left settings.
Let's say: after the show ip route
command, the problematic 92nd displays Interface vlan5 , and on the analogue Interface vlan5-0
Although in all other places VLAN5 is displayed the same way.

As a result, we have the absence of registered gateways and, possibly, network segmentation, as nApoBo3 wrote in the comments.

Answer 1

Let's start with a simple poppy on the port of the problematic vlan? Where do you do ping from the access switch? Does he have a terminal of several vlans?

Answer 2

You wrote that the switch is reporting a POE problem.
Ring the cable, connect another device with poe on the other side of the switch to check
Bring the non-working device to the switch and connect it with a short patch cord ... The laptop works, because it is not checked whether
the poe is “alive”