Docs

Backup alerts and troubleshooting

The alerts backup raises, what raised each one and how to retry from the alert, when they clear, what stays quiet on purpose, and how to fix failed, overdue, helper, interrupted, slow and disk space problems.

Backup problems show up in two places: Needs attention on the Backup overview, and as alerts in the Alerts area. Backup alerts go through the same alert handling as everything else, so your alert rules can turn them into tickets, and they clear themselves when the problem is fixed.

Needs attention

The Needs attention panel on the Backup overview lists what needs a person, most serious first.

Needs attention with failed backups and unprotected devices
Needs attention with failed backups and unprotected devices
  • Failed: a backup that failed, with the reason in plain words and the next step (the helper stopped reporting, a full disk, a VSS snapshot that failed, "Mailbox over its Microsoft quota"). Technical details under it shows the raw text. See the log opens the job that failed, scrolled to the device's volume or the person's part that failed, with its details open.
  • Nothing selected: a device with protection on but no volume, virtual machine or folder ticked, so no backup runs. Choose what to back up opens it.
  • Overdue: a device or Microsoft 365 item with no good backup for longer than its threshold.
  • Not protected: devices with the agent that have no backup. Review opens them in the Devices list so you can protect the ones that need it.
  • Helpers that will not start, tenants that refuse sign-in, and the backup server running low on disk space.

Above the list, a line says how the items stand in Alerts: how many are open alerts, how many failed once and will alert only if their next run fails too, and how many are only overdue by their schedule. Open backup alerts opens Alerts filtered to backup, so the two lists count the same problems. An item with an open alert has an Alert link to it.

Paused devices, backups that completed with a few skipped items, backups interrupted by a restart or an update, and servers used only as sandbox hosts are not problems, so they are not listed.

Show all lists the rest.

The alerts

Open Alerts and choose Backup under By source in the sidebar (or set the Source filter to Backup) to see only backup alerts.

Backup alerts in the Alerts list, with Backup chosen under By source
Backup alerts in the Alerts list, with Backup chosen under By source

Every backup alert opens with Raised by: the device and volume, the tenant and part, the repository or the boot check it is about, as a link, with whether it is still true and its key facts, such as the job that was running. Where a retry can fix it, the alert offers to back up again from the alert page, with a preview of what will run. See What raised an alert, and fixing it from the alert.

Problem Raised when Clears when
A backup failed A volume, folder or guest failed twice in a row (one failure that its retry fixes tells nobody). Critical The next good run of that unit
Disk read error A device backup hit an unreadable area of a disk, for example "E: on SQL-01 has an unreadable area (disk read error at 417 GB). The disk or its datastore may be failing. Check the host storage, then run chkdsk." Critical at once, and the device's health turns Critical The unit backs up cleanly
Never backed up A selected volume, folder or guest still has no complete backup after its first scheduled run and one retry, with the last failure in plain words. Critical The first good backup
Microsoft 365 backups failing One alert per tenant naming each person, site or team whose part failed or finished with errors twice in a row, for example "Microsoft 365 backups failing in Contoso: 2 people and 1 team". Critical while any part failed outright, a warning when they only finished with errors. Microsoft slowing a backup down for a run never alerts No part is left on the list
Overdue No good recovery point for longer than Overdue after (default 48 hours for devices, 24 for Microsoft 365) A good recovery point lands
Helper problem The backup helper keeps failing to start or update on a protected device The helper runs normally
Sign-in refused Microsoft refuses a tenant's credentials, or consent was withdrawn. Critical The next accepted sign-in
Storage Backup storage cannot be written, or a folder has less than 10% free The next storage check passes
Backup repository damaged Stored data in a repository could not be read back. One alert per repository, Critical Nothing in it is left damaged (repaired or healed)
Backup repository maintenance falling behind Any repository has more than 500 index blobs, no quick maintenance for 48 hours or no full maintenance for 10 days. One warning for the install, naming each They have all caught up
First backup slow A tenant's first full backup runs slower than its estimate, with why. Sent to whoever connected the tenant It speeds up or finishes
Boot checks cannot run A sandbox host cannot run boot checks (it is unreachable, or the port it needs is closed). One alert per host, naming the host and the port The host can run checks again
Skipped items More than a few items of a part are skipped, past the thresholds in Settings > Backup > Alerts. Accepted and excluded items are not counted. A warning first, an alert the next run The part is back under the thresholds
Interrupted again and again A part was interrupted 3 times in a row before it finished (warning, with the last reason and what is stored). Restarts and updates of Tenvara itself are not counted The part completes
Slow download A OneDrive, site or Google Drive backup stayed slower than 256 KB/s for an hour (warning, with why and the job it happened in). Mail, calendar and contacts are not measured this way The part's next run is not slow
Backup server disk space The backup server's scratch, data or database disk is below the line, so large files are paused. Warning, critical below half the line There is room again
Local copy A device's local copy has had a problem for 24 hours. Backups still reach the server The local copy is healthy or seeding again
Second copy behind A repository's second copy is further behind than Alert when a copy is behind by The next good copy
Test restore or boot check failed A scheduled test restore failed, or a machine did not boot the set number of times in a row The next pass
Restore search is behind A backup that search should have read has waited longer than Alert when search is behind by. The alert is kept up to date while it lasts Search has caught up
Possible ransomware Comparing two recovery points found more than the set share of files changed or deleted, when Raise an alert too is on You resolve it once you have checked

Alert titles say what failed, for example "Backup of D:\Shares on HVD-DC01 failed" or "Microsoft 365 backup failed: Accounts (mail)".

Each alert belongs to the customer and, where there is one, the device, so it shows on their pages too. The storage alert belongs to no customer.

Stopping protection of a Microsoft 365 item clears its alerts. Pausing a device or customer clears its overdue alerts and holds new ones until the pause ends. Retiring a device clears every backup alert it had, its boot check alerts included. Overdue checks run every 15 minutes.

What never alerts

  • A backup cut short by a restart or update of Tenvara. It is interrupted, carries on by itself and is not counted as a failure. See Backups through restarts and updates.
  • Items that are behind only because backups were deliberately off (the backup service was stopped, the Backup module was switched off, or the customer is in takeover mode). They show Paused and catch up at their next backup.
  • A server used only as a sandbox host. It has no schedule, so it is never overdue.
  • A boot check that could not be carried out, for example because the sandbox host itself stopped the machine. It is recorded as not checked, with the reason, and runs again later.

You can switch a kind of backup alert off altogether, or for one customer, in Alerts > Alert types. See Alert types.

A device's Backup section also warns, without an alert, when a SQL Server log is growing because nothing truncates it. See Application-aware and SQL Server backups.

Fixing common problems

A device backup failed

The reason on the item says what failed in plain words and what to do next. When it says a disk has an unreadable area, treat it as a failing disk or datastore: check the host storage, then run chkdsk (Windows), smartctl and fsck (Linux) or First Aid in Disk Utility (Mac), before backing up again.

  1. Open the job from See the log or the device's Recent jobs and read the log. The last lines usually say what went wrong.
  2. Check the device is online and has free disk space. If the job says "Waiting for the device to come online", it has not failed: it starts when the device is back.
  3. Press Back up now on the device to try again once it is fixed.

"Could not create a VSS shadow copy"

Windows could not take a snapshot of the volume. Common causes are too little free space for the shadow copy, another backup product holding VSS, or a VSS writer in a failed state.

  1. Run vssadmin list writers on the device (through Remote control or a script) and look for writers that are not Stable.
  2. Restart the service behind a failed writer, or restart the device.
  3. Make sure no other backup software is running at the same time.
  4. Back up again.

A device is overdue

The device has not had a good backup for longer than Overdue after.

  • It is off or offline at the scheduled time: common with laptops. A backup still queued for an offline device when its next backup falls due ends as Missed, device offline, and the next one takes its place. Move the schedule to a time it is usually on (see Schedules and retention), or use Back up now when it is next online.
  • A job is stuck in the queue waiting for the device: it starts when the device comes back.
  • The helper is not running: see below.

The helper will not start

The device page shows the helper's state under Helper, and the Devices list has a Helper column.

  • Backup not supported: no helper build exists for the device's platform. Check Status in backup settings (see Setting up backup).
  • "helper download failed": the agent could not fetch the helper. Check the device can reach your Tenvara address. It retries on its own, backing off up to every 30 minutes.
  • "helper keeps exiting": the helper starts and stops, and the message after it says why (for example a VSS error, below). If a newly updated helper fails to start three times, the agent goes back to the previous version automatically. The helper's own log is helper.log in the agent's backup folder on the device.

Microsoft 365 parts failing

  • Sign-in refused on every part: the tenant needs consent again. See Microsoft 365 backup.
  • Only Teams failing: the app has not been approved for Microsoft's protected Teams API yet.
  • A user's OneDrive is skipped with "no OneDrive provisioned": the user has never opened OneDrive, or is not licensed for it. This is a skip, not a failure.
  • A large file that changes while it is copied: a file whose version keeps changing because it is always open or syncing, such as a running virtual machine's disk or an open Outlook data file, cannot be stored consistently. From the third run in a row the skipped item says so and what to do: move it out of OneDrive or SharePoint, or exclude it. A file Microsoft lists at a different size from the one it serves is not a change: it is copied at its real length.
  • Throttling: Microsoft limits how fast any app can read a tenant. Tenvara paces itself to the tenant's SharePoint requests a minute, backs off when Microsoft asks it to, and carries on from where it stopped in the next run. A very large first backup can take several runs. See Resuming large and interrupted backups.
  • A few items skipped: a handful of emails or files Microsoft will not hand over is normal and raises nothing. Open the job's skipped items and Why is this?. Temporary errors from Microsoft (500, 502, 503, 504, 429) are retried in the run and on later runs. A file Microsoft's malware scan flagged can never be copied: the item offers Open the file's location.
  • The same items skipped every run: once you have looked at them, Accept them, or Exclude an email or file from its mailbox's, OneDrive's or site's backups. They leave the job's warning line and are listed under Known skips (N); only a new or changed skip brings the warning back. See Skipped items.
  • An email Microsoft cannot export ("MIME content conversion failed"): nothing to do. It is backed up as a copy rebuilt from Microsoft's message data, and counts as backed up. See Searching and browsing backups.
  • A first backup that does not start: open the first backups queue. A held tenant says why (takeover mode, Backup switched off, consent, the connection, the SharePoint allowance, low disk space or a person's hold) and offers the fix.

A backup keeps being interrupted

Open the part's unfinished run on the device's or tenant's page: it says how often it was interrupted and why.

  • The device restarted or went to sleep: common with laptops that sleep overnight. Each run keeps what it stored, so the backup finishes over a few nights. Move the schedule to a time the device is usually awake to finish sooner.
  • The backup helper was stopped: the agent was updated or the device was shutting down. Nothing to do; the next run carries on.
  • The backup service stopped: the backup server restarted, for example for an update. Interrupted parts carry on as soon as it is back, and these interruptions are not counted towards the alert.

A cloud backup is slow

The slow warning says why: throttled by Microsoft or Google (raise SharePoint requests a minute per tenant only for a customer with more licences), the network, storage (writing into your backup storage), the processor of the backup server, or the server-wide limit on files at once. When it says index maintenance needed with a number of index blobs, the repository's index has grown too big: press Compact the index now on the repository in Backup > Repositories (see Repository health, healing and verification). See the Cloud downloads settings in Resuming large and interrupted backups.

The backup server is low on disk space

Large files pause while the backup server's own disks are below the line shown under Backup server disk space on the Defaults tab of Settings > Backup. Everything else carries on, and the paused files are fetched once there is room. Free space on the disk named in the alert, or give the server a bigger disk.

Storage alert

  1. Open Settings > Backup, the Storage tab, and press Check on the location to see the error.
  2. For a folder, free space or move to a larger disk (copying the repositories across first).
  3. For S3, check the keys, bucket and endpoint still work.

Backups that could not be written will run again on their next schedule once storage is fixed.

Keeping on top of backup

  • Look at the Backup overview each morning. Success, last 24 hours should be close to 100%.
  • Set an alert rule that turns critical backup alerts into tickets, so failures get an owner.
  • Review the Not backed up view of Backup > Devices weekly: new servers in particular should never be left without a backup.
  • Turn on "Morning backup summary" in your notifications for one email a day with every failure and overdue item. See Setting up backup.
  • Test a restore from each customer regularly (see Restoring).

Was this page helpful?

Thanks for the feedback.