Added
- A safety stop for mass upgrades. Settings has a new Safety stop section. Set how many devices may fail in one upgrade run, or what share of them, and once that is reached mikr starts no further devices: the ones already upgrading finish and the rest are held. It applies to upgrades started from the table, to the queue and to scheduled upgrades, and both the upgrade table and the queue take a one-off limit for a single run. When a stop trips you get a warning, and scheduled upgrades pause until you press Resume on the Schedule tab, so a release that has just failed does not go on to more devices at the next scheduled time. Everything is off until you set a limit. Reported on GitHub (#80).
- A safety stop for Mass Commands and templates. The same section sets a limit of failed devices for Mass Commands and for running or deploying a template: past it, the devices not yet reached are held and left untouched. You can also have mikr wait a few seconds after each device and check that it still answers — a device that stops answering counts as a failure, which catches the command that cuts a router off before it reaches the next one. With either option on, devices are handled one at a time. Both can be set for a single run under Safety stop (this run only). Reported on GitHub (#81).
- Pin a scheduled upgrade to the version you reviewed. A scheduled upgrade used to install whatever the update channel offered when the time came, so a release published after you scheduled it went out without anyone looking at it. Tick Pin to the version available now when scheduling — on the Schedule tab, from Schedule… above the table, or on a queue preset — and at run time a device is upgraded only if the version on offer is still the one you saw. If a newer one has appeared, that device is held and the log says which version was pinned and which was offered. A device that has never been checked for updates cannot be pinned, and mikr tells you which one. Reported on GitHub (#82).
- One filter for every place you pick devices. The Upgrades table, the queue, the schedule picker, Mass Commands and the template run dialog share a filter box. A plain word matches name, identity, address, site or model;
name:,host:,site:,model:,ros:,fw:,channel:andtag:narrow by one field;status:behind,fw-behind,current,unchecked,onlineandofflinepick by state. Terms combine, and values are regular expressions, soname:^cap site:officedoes what it reads. Hidden devices are unticked, and Select All and the site and tag shortcuts act only on what is shown. The queue also gains Add all shown. Reported on GitHub (#72, #79). - Device names are links. On the dashboard, the Devices and Sites pages, the Upgrades page, the device page and the Security page, a device name can be middle-clicked, Ctrl/Cmd-clicked or opened in a new tab from the context menu. Reported on GitHub (#70, #74).
- Sites and devices are listed in a sensible order everywhere. Site lists followed whatever order the devices happened to be stored in. Sites and the devices in them are now sorted by name on every page, the way a person would sort them:
2before10, and capitals mixed in with lower case. Reported on GitHub (#71).
Fixed
- History charts were drawn from every sample instead of averaging. Charts over 24 hours, 7 days and 30 days are meant to combine samples into intervals, but plotted each one individually, so a 30-day chart loaded more than 7,000 points per metric where about 120 are needed. They now average as intended and load much faster. This covers device health, optical SFP readings, BGP and interface traffic.
- The Activity log missed real actions and recorded a page view. Opening the Activity page added a "Viewed users list" entry of its own — and for an administrator who is not a superadmin, a failed one, with an empty user filter. Several actions were never recorded at all, among them network scans, backup exports, cancelling an upgrade queue, saving settings, changing a device's password and signing out. These are now recorded, along with upgrade schedules and presets, template runs, Wi-Fi password and LTE firmware changes, and clearing collected logs. Each mass upgrade also ends with an entry giving how many devices succeeded and failed, including runs started by a schedule, and a queue that finished with failures is marked as failed rather than completed. Reported on GitHub (#77).
- Version colours on the Upgrades page did not match what needs upgrading. A firmware version that was behind was highlighted but a RouterOS version was not, and after an upgrade the highlight stayed on a version that was now current. Both now highlight when behind and clear once the device is up to date. Reported on GitHub (#75, #76).
- Dragging a device into the upgrade queue only worked at the very bottom edge. You can now drop it anywhere on the list — it goes in where you let go — or on the strip below the list to add it at the end. Reported on GitHub (#78).
- The memory, database size, uptime and device count at the bottom of the menu stopped updating after the page loaded. They now refresh every minute. Reported on GitHub (#73).
- A device added by a name that has only an IPv6 address could not be monitored over SNMP, and showed as not accessible although SSH to the same name worked. mikr now uses IPv6 for such a name. A name that also has an IPv4 address keeps using IPv4, so nothing changes for dual-stack devices. Reported on GitHub (#83).
- WireGuard transfer counters were shown far too small — 342.9 MiB appeared as 342 B. They now show the same figures as the router.