SSH continuously taking up 25% CPU on a CCR2004, stays this way even when SSH service is disabled. Also rebooting the router which used to be very quick now takes too long, as if a service is stuck, blocking the shutdown phase:
It is actually returning back to /ip firewall connection tracking set enabled=yes instead of /ip firewall connection tracking set enabled=auto.
Yes, I have reported this exact issue to Mikrotik support. It was not always reproducible. But I have received an email from their bugtracker yesterday saying it was resolved. So apparently they've fixed something.
What's new in 7.26beta4 (2026-10-27):
- app - add remove-certificate command for certificates not used by any app;
- bgp - improve update processing performance when multiple peers are used;
- bonding - add lacp-strict setting for 802.3ad bonds;
- bridge - fix incorrect root-path-cost calculation in MLAG when the root bridge is on a single-connected port;
- bridge - fix loop-protect not working when a VLAN interface is added to the bridge (additional fixes);
- certificate - add option for pasting certificate in PEM format in add and import commands;
- certificate - fix CRL download failing when the server responds with an HTTP 3xx code;
- certificate - fix empty error message when ACME DNS validation times out;
- certificate - fix services continuing to use the previous certificate after ACME renewal;
- cmr - fix configuration display not showing effective values for unset fields;
- cmr - fix VLAN rules being applied to all ports when port-labels is not set;
- cmr - improve stability;
- cmr - replace dashboard value bars with grouped bar graphs;
- cmr - require write policy to run the alert test command;
- console - fix custom SSH client login parameters being ignored (introduced in v7.26beta1);
- console - fix deprecated argument names not being readable with get and proplist;
- console - improve stability;
- disk - fix NFS clients being served the wrong export after a reboot;
- ethernet - disable EEE on RB4011, RB1100AHx4, L009, L11, L23 devices;
- ethernet - do not allow adding duplicate RPS interface entries;
- ethernet - fix broadcast traffic still reaching the CPU when port isolation is used on Atheros and MediaTek switches (additional fixes);
- iot - fix LoRa channel configuration changes failing with a validation error;
- ipsec - fix XFRM interfaces staying down until a reboot;
- ipv6 - fix from-pool address lifetimes not updating after a prefix renewal (introduced in v7.25);
- l3hw - improve stability on CRS8xx series switches;
- log - fix logging actions with target=script not running the script after a reboot;
- lte - fix eSIM provisioning from the mobile app (introduced in v7.26beta1);
- lte - fix Telit LE910C4-WWX modem serial driver not binding at cold boot;
- lte - fix traffic not passing with multiple locally terminated APNs for R11e-4G modem;
- lte - reduce eSIM pending notification send retry count to 2;
- netwatch - fix check-certificate not working for HTTPS-GET probes;
- netwatch - fix wrong checksum in the first ICMP echo of a test on TILE devices;
- poe-out - fix missing status column in print output;
- ppp - fix BG770A modem not initializing after a firmware upgrade until a reboot;
- route - add RPKI ASPA validation support;
- route - fix routing memory statistics missing some memory pools;
- sfp - fix DAC EEPROM corruption when disabling 400G ports on CRS804 and CRS812 switches;
- socks - fix comments not being saved;
- ssh - refactor SSH internal processes;
- switch - add a high-priority CPU queue for control traffic on Marvell Prestera switches;
- switch - fix CPU-sent VLAN-tagged packets with the DEI bit being dropped on Marvell Prestera switches;
- system - improve stability;
- tr069 - add nodes for GPS information;
- usb - add overcurrent monitoring and per-port power cycle for hAP be3 Media;
- user-manager - fix database not opening after reboot when db-path is several directories below the mount point;
- webfig - fix missing default values for CMR label fields;
- webfig - fix the Power Cycle action not working;
- wifi - fix virtual interface restart loop when reselect-interval and reselect-time are configured (introduced in v7.25);
- wifi - initial easymesh controller and agent implementation (CLI only);
- wifi-qcom-be - fix station mode not connecting when limited to an older standard than the AP;
- wifi-qcom-be - fix station not connecting when limited to narrower channel than AP;
- winbox - allow adding LTE interfaces with a master interface and APN profile (additional fixes);
- winbox - make RPS menu configureable;
I don't know why, but LAG (on two ports) between hap be3 media and the proxmox lab server became uncreatable. All links were up, but without traffic, even ping didn't work.
Revert back to 7.25 and it works.
After upgrading my hAP be3 Media from 7.26beta3 to 7.26beta4, devices on ether2-ether5 can no longer connect. Setting hw=no on these ports fixes it.
Hi,
after the upgrade from 7.26b3 to 7.26b4, printing ipsec policies on the console sometimes needs a second attempt, the first attempt shows no results:
[admin@mikrotik2] > /ip ipsec policy print
[admin@mikrotik2] > /ip ipsec policy print
Flags: T - TEMPLATE; * - DEFAULT
Columns: PEER, TUNNEL, SRC-ADDRESS, DST-ADDRESS, PROTOCOL, ACTION, LEVEL, GROUP
# PEER TUN SRC-ADDRESS DST-ADDRESS PRO ACTION LEVEL GROUP
0 T* ::/0 ::/0 all default
1 site-2-site yes 192.168.200.0/24 192.168.100.0/24 all encrypt unique default
Kind Regards
I see a lot of improvements for be3. But is ipq-ppe based devices still being working on ? We still don't have proper hw offloaded bridge support for ax3 for exemple...
Thanks ![]()
I suppose it's "only" for wifi-qcom-ac, wifi-qcom, wifi-qcom-be and wifi-mediatek
I tried this version but got no internet.
The static ip address I assigned is not getting picked up. It used the ip addresses assigned by the netinstall package.
Fetch crl's stopped working too.
Didn't go any further and went back to 7.24.5.
I don't see any problem with ports on both be3 with 7.26beta4. Both devices are running default configuration.
It is located /interface/wifi/easymesh
After removing a device from the CMR, the CMR Manager has generated an autosupout, a ticket has already been created.
This hasn’t happened for me on beta1 or beta3 so far.
I prefer the look of the CMR interface in WinBox; if possible, they should be identical.
Furthermore, the number of radios is listed under ‘Access Points’, so I think the labelling is incorrect.
Yes, I get that too. I'm on Mac, but happens on Linux too. Seemingly anything with CMR. So I think there is bug, but maybe we need a new WinBox to get that fix/icons, IDK.
Icons and error message fixed in WinBox 4.5rc1.
Thank you, working.![]()
(post deleted by author)
If you like the CLI: the cmr dashboard looks awesome!
/cmr/device/dashboard
Me too.
After disabling hw offloading, it is ok.
/interface bridge
add admin-mac=38:32:7A:XX-XX-XX auto-mac=no name=bridge1 protocol-mode=none
/interface bridge port
add bridge=bridge1 hw=no interface=ether1
add bridge=bridge1 hw=no interface=ether2
add bridge=bridge1 hw=no interface=ether3
add bridge=bridge1 hw=no interface=ether4
add bridge=bridge1 hw=no interface=ether5



