Skip to content

TasmoFleet 0.42.0 "Cape Otway"

Choose a tag to compare

@github-actions github-actions released this 20 Nov 08:46
· 50 commits to main since this release
Immutable release. Only release title and notes can be modified.
v0.42.0
e85533c

Upgrade notes

  • Migration 0054 adds the columns more_group_topics, more_group_topics_read_at and more_group_topics_gen to devices (GroupTopic2 to GroupTopic4 of each device, #1005). No data is changed.
  • Right after the upgrade, GroupTopic2 to GroupTopic4 are unread for every device. Like any unread group topic, that holds back writes and transfers over MQTT for the whole fleet until TasmoFleet has asked each device once (one device every 2 seconds, so about a minute and a half for 45 devices); in that window they go over HTTP where HTTP reaches the device (#1005, thanks @Noschvie).

Added

  • Bulk result: a click on the Succeeded, Failed or Skipped tile shows only those devices (a second click shows all again), the tiles stay in view below the header while the list scrolls, and every device in the list links to its own page (#998, thanks @Noschvie).

Changed

  • Updates > Transactions, and the Dashboard card of the same name, are now titled "Transactions" (they list every bulk transaction, not only updates). The status bar says "Scheduled backup", because it shows the last scheduled run only and a "Back up" from the device list is a transaction of its own, and it counts like the Last run card: "36 backed up · 5 failed · 3 skipped", three numbers that add up to the devices of the run instead of "36 of 41 backed up" (#992, thanks @Noschvie).
  • Backups: every backup run now starts its devices in name order, A to Z (case-insensitive, then by device ID), the same way the scheduled run always did. "Back up all now" used to start them in the order the devices were added and "Back up selected" in the order of the device list. The result page of a backup run lists the devices in that start order (the run still does two at a time, so the finish times can differ slightly), and the run's rows and the Backups table show the device ID as a small muted "#N" next to the name (#993, thanks @Noschvie).
  • Backups: the Last run card is now its four tiles (Started, Backed up, Failed, Skipped) and no longer lists the failed devices again, with "and N more" and "Show all", or the skipped ones. The devices with their reasons are in the table below, which the Failed and Skipped tiles open filtered. The note that skipped devices are not failures and not in the notification now sits on the Skipped tile (#996, thanks @Noschvie).
  • The Identity card shows the TasmoFleet ID with a copy button, and the log line of every audited action on a device carries the device as "device":<id> like all other lines, so one filter finds everything about a device (#1004, thanks @Noschvie).

Fixed

  • Device page: the Policy tile counts only the rules that apply to this device (not every rule in the fleet), and a device that no rule applies to says "No rules apply to this device" instead of "2 / 2 Matches all rules" (#986).
  • The device page, the update preview and the first line of an update's log now word an update's path the same way, and no longer say the firmware download goes over MQTT: only the commands (OtaUrl and Upgrade) go over MQTT, the device fetches the image over HTTP from TasmoFleet's firmware server. The path's badge on the device page names the commands' transport ("Commands over MQTT" or "Commands over HTTP"), the note under it and the preview row say both parts ("commands over MQTT, the image over HTTP from TasmoFleet's firmware server"), and the preview now shows this note for the HTTP paths as well (#987).
  • Saving a device without changing any of its fields (for example when only the MQTT password mode changes) no longer writes the device again or adds an empty device.update row to the audit log; the password-mode entry stays, and a device without tags that is announced again unchanged no longer adds one either (#988).
  • Run now and Check now say when the per-device read cooldown (5 minutes) left a device alone and from when it can be read again, and the cooldown counts only a read that delivered every setting the device's rules judge: an MQTT message, a Make compliant check or a half-failed read no longer holds a check back or pushes it back, so a rule no longer stays "Not judged" because of an unrelated read. A device that never returns one judged setting is now read on every Run now (at most once a minute) instead of every 5 minutes (#989).
  • The save bar of System, Settings and System, Snapshots now shows only while the form has unsaved changes, as it already did on the Backups page. Before, it stayed on screen all the time and covered the lower part of the Device MQTT passwords card on the MQTT tab. After a refused save, a step-up request or a test answer without script the bar stays on screen, because the page then shows values that were not saved (#990).
  • Backups: the answer to Back up now and Retry (a refusal during the cooldown, a failed backup, the API turned off, a missing MQTT password) now appears directly under the bar of the Backups tab, under the empty card when there are no backups, and the page scrolls to it. With a long list it used to land below "Recent restores", off-screen. The wait in a refusal rounds up to the next full second, so it never says "possible in 0 s" (#994, thanks @Noschvie).
  • Backups: a device that only MQTT can back up (no answer over HTTP, for example in another LAN) and whose MQTT password is not set or was refused is no longer sent a backup over HTTP that can only time out with "The device took too long to answer". The backup ends at once with "Needs an MQTT password", for Back up now, Retry, Back up selected, Back up all now, the scheduled run (counted as failed, so it stays in the notification) and the backups before an update, of a template job and of Replace. When TasmoFleet knows of no recent HTTP answer, it asks the device once over HTTP (Status 11) first. A device that answers over HTTP keeps backing up over HTTP without the password, and when such a backup fails, the error now also says why MQTT could not take it. The Backups table, the device's Backups tab and the result of a backup transaction link to the device's MQTT password form, and with several such devices the Backups page links to Device MQTT passwords on System › MQTT. A backup transaction's result now shows the same reason sentence as the backup history instead of the internal error text (#997, thanks @Noschvie).
  • Console: a Status 0 typed for a device reached over MQTT only is refused right away with the reason (Status 0 never goes over MQTT, and the device's HTTP is not known to work) and the hint to ask Status 1 to Status 11 instead, instead of being sent over HTTP and ending in "The device did not answer in time"; nothing is sent. Over a working HTTP it is sent as before after the confirmation. A bulk Status 0 lists such devices as "Not sent" and leaves them out (#1001, thanks @Noschvie).
  • Make compliant page: a device that deviates from Advisory rules only is no longer counted as "0 devices" or blocked by a hint; the title names the one selected device (or counts the selected devices), the subtitle says "Nothing to run yet" while no command is planned, Update plan is enabled only while a tick differs from the drawn plan, and Cancel returns to the device page the plan was opened from (#1002, thanks @Noschvie).
  • After a firmware update done outside TasmoFleet (for example OtaUrl and Upgrade 1 in the console, which on an ESP8266 goes through the minimal image), the device page shows the new version and OTA URL within a minute of the boot instead of up to an hour later: a restart read that met the minimal image no longer counts as the restart's read, and a live INFO1 or a backup that names another version, or a console write the device accepted of the OTA URL, a group topic, the topic, a name or the hostname, now reads the identity at once (nothing of the message, backup or answer is stored without the identity read's checks). A device in a boot loop on one version is still read about once an hour. The Identity card now refreshes live like the header when what it shows changed (#1003, thanks @Noschvie).
  • A restart read whose identity check was refused because another device answers at the device's address no longer lets every further restart read the identity again: the next restart's identity and data read waits 10 minutes after the refusal. Restart messages can be published by any client on the broker, so a refusal that lasts no longer turns each of them into a full identity read. The first refusal by minimal firmware in an hour does not wait, so the boot into the new firmware after an update is still read in full at once (#1003).
  • The check that keeps a write or a transfer off MQTT when another device could receive it now knows all four group topics of every device, not only GroupTopic1: TasmoFleet reads GroupTopic1 to GroupTopic4 with one bare GroupTopic (MQTT first), stores them, and the device list's Group topic column shows the first one with "+N" for the others (all of them on hover), the list filter finds a device by any of them, and the Identity card lists them all. After TasmoFleet itself sets a group topic (console, bulk command) or restores a configuration, the device's group topics count as unread until they have been read again, a few seconds later (#1005, thanks @Noschvie).
  • Refreshing a device's identity after several refreshes in a row no longer waits behind a pile of identical background reads of the network facts: a device waits for at most one such read (#1015).