Strange, on SWSRV01 and SWSRV02 (both HP 2920-24G) there are few non Default (port based, since the no ip address) VLAN IDs configured (VLAN IDs are: 100, 110, 120, 130, 140, 150, 160 and 170)...but NO port of those Switches is assigned as tagged member (of one or multiple) or untagged member (of one and only one) of those VLAN IDs (the directive on VLAN ID 1 means you have all ports - 1-24, A1-A2 and B1-B2 of each 2920-24G - as untagged members of Default VLAN ID 1, that's OK...but what's the purpose of having others VLAN IDs defined if no physical/logical port is configured to be member of them?).
AFAIK with STP properly configured on all four units you should see at least two uplinks automatically down (respectively one of those originating on SW03 and originating on SW04) facing the 2920s...that's to avoid a double loop (SW04-SWSRV02-SWSRV01-SW04 and SW03-SWSRV01-SWSRV02-SW03).
I'm not an HPE Employee
If you are going to deploy backplane stacking on two HP 2920 (using dedicated Stacking modules and Stacking Cables) your network topology is going to change too: at that point the cluster of two stacked 2920 will become a single virtual logical switch and both SW03 and SW04 can then be connected using, each one, a Port Trunk (where Port Trunk means here aggregated interfaces also known as LAG or BAGG, configured to use LACP) with member links (e.g. SW03 Trk1 Ports 23+24 and SW04 Trk1 Ports 47+48) terminated and distributed on the stack (let me say SW03 Ports 23+24 on SW01 Port 23 and SW02 Port 23, SW04 Ports 47+48 on SW01 Port 24 and SW02 Port 24) for balancing in case of Stack node failure.
Clearly also on the Stack you should configure Port Trunking groups (Trk1 Ports 1/23+2/23 and Trk2 Ports 1/24+2/24) with LACP.
No more physical loops (It's a simple star topology), STP enabled.
I'm not an HPE Employee
Shouldn't you use basically RSTP (Rapid STP) globally on all your Switches? why citing MSTP (Multiple STP)? is that necessary to manage different STP regions related to different VLANs by adopting MSTP or using, as it seems you're already doing now on SW01 and SW02, the RVSTP (Rapid per-Vlan STP)?
What exactly you're trying to achieve (or to avoid) in case the network topology will remain as is...with various Port-based or Protocol-based VLANs you're going to deploy?
RSTP, AFAIK, should be supported on HPE OfficeConnect 1910 and on HP ProCurve (now Aruba) 2920...3Com should be checked, but should be easy enough.
Edit: 3Com Baseline Switch 2952-SFP is basically a 3Com branded 1910-48G with 4 additional SFP...and it supports RSTP for sure.
I'm not an HPE Employee
fga76 wrote: I was think about my network layout. I do not have an uplink between SW03 and SW04.
Exactly, you haven't an uplink between SW03 and SW04...if an uplink ever exists it will produce another loop other than your two loops (SW03-SVR01-SVR02-SW03 and SW04-SRV02-SRV01-SW04). So it's good that you haven't that additional uplink.
Should I have an uplink between those two switches?
No.
And then disconnect SW03 from SRV02 and SW04 from SRV01?
Eureka, no more physical loops: your topology will became a daisy chain SW03-SRV01-SRV02-SW04.
The path from SW03 to the servers on SRV02 would go through SW04. And vice-versa.
Yes, that's true...such design could (or couldn't) became a bottleneck, it all depends where you client hosts are and where your server hosts are and where the majority of data traffic flows.
For some reason, when the ports within those 4 switches are in trunk mode, I got a loop.
IMHO you got a loop in any mode because there are two loops (see above). At least if you haven't configured something like MSTP or PVSTP correctly instead of (R)STP.
Or disable STP on all switches, create the trunks in the current layout and then reestablish STP?
Or rethink your network topology in order to prevent loop forming...and then enable (R)STP.
I just don't get why the loop when in trunk mode.
In Trunk Mode only? Are you sure? I ran out of ideas.
I'm not an HPE Employee