Last Week on My Mac: Giving a fsck

Sometimes Macs behave in ways that you don’t forget. It was nearly six years ago, in 2019 when my iMac Pro was still running macOS Mojave, that it took over half an hour to start up in Safe mode. As became obvious, that was because it checked the integrity of every snapshot, including its long series of Time Machine backups. That behaviour stopped with the release of Catalina later that year, and revisiting the experience last week was interesting, when I examined what Safe mode does now in Sequoia.

Safe mode fsck

Checking disk integrity has long been a feature claimed of Safe mode, although Apple has only recently clarified that its checks are “similar to the more comprehensive check performed by the First Aid feature of Disk Utility,” unlike those in the days of Mojave.

Because these are run when starting up into Safe mode, macOS can’t report them directly to the user. There are only two places for their results to be recorded: in the fsck_apfs log at /var/log/fsck_apfs.log (and its error log at /var/log/fsck_apfs_error.log), and in the Unified log, neither of which are likely to be visited by users. It doesn’t seem possible that checks could be run before entries are made in the Unified log, and if they were their results would be inaccessible, and of no value to the user.

If you do look there, you might be surprised to discover that the checks run by fsck_apfs are ‘quick checks’, described as a “check whether the device is `clean’. If device is an APFS volume, fsck_apfs will quickly check the APFS container and the specified APFS volume. If device is an APFS container, fsck_apfs will quickly check the APFS container and all the APFS volumes in it. By default, no repairs are attempted during a quick check.” Thankfully they also no longer include Time Machine snapshots or backups.

For users, the most important of volumes to be checked is the Data volume in the current boot volume group. Yet careful examination of both log records reveals that such checks fail, returning error 65, because the device is mounted with write access. Although they could be performed using fsck_apfs‘s live verification, no attempt is made to do that. From the active boot volume group, only the Recovery and Update volumes are checked, together with all accessible external volumes.

Normal user fsck

Before you think that’s better than making no checks at all, the sequence of fsck_apfs checks run at the start of Safe mode appears to be identical to those made during startup into normal user mode, and reports the same failures. Thus, in this respect Safe mode is no different from normal, despite Apple listing this as a feature of Safe mode.

It’s worth reading the fsck_apfs man page to see the many options available in this command, which is the basis for all checks and repairs made to APFS volumes and containers. There are two in particular that you may prefer to invoke rather than using Disk Utility’s single option to both check and repair. My favourites that make it worth opening Terminal instead of Disk Utility are -n to prevent it from attempting to make any repairs, and -S to skip snapshots.

Given that fsck_apfs appears unable to repair any errors it finds in snapshots and admits to that in its man page where it states that “no repairs can be made”, it puzzles me that without the -S option it will still check all snapshots it comes across when running its checks. Perhaps it’s time to alter that default behaviour and make checking snapshots the explicit option.

First Aid

Disk Utility’s First Aid is descended from a succession of utilities starting with the Disk First Aid app in Classic Mac OS, which separated disk Verify and Repair features.

Mac OS X merged the two Classic apps Drive Setup and Disk First Aid, but kept them as distinct tools within its new Disk Utility app and retained separate Verify and Repair features.

Those two features were still present in Mountain Lion or Mavericks, shown here, but soon afterwards only Repair was available.

Now that APFS is mature and has proven itself more than a worthy successor to HFS+, the need to both check and repair its containers and volumes should be abating. This would be a good time to return to separate controls, allowing users to check APFS in the expectation that there are no errors to be found. This is more important now because of the absence of any alternatives.

Monopoly

Disk Utility is in a unique position. Apple has chosen to lock third-parties out of developing competing disk repair utilities, in particular those that can be used in Recovery mode. Users can’t opt for apps with more or better features, putting the onus on Apple to provide a comprehensive solution that meets users’ wishes, not its idea of essential requirements.

Conclusions

Apple needs to be explicit about what disk checks are run in Safe mode, and where users can inspect their results.
fsck_apfs is flexible and powerful, but has strange defaults that merit reconsideration.
Disk Utility should offer more options beyond its single ‘check and repair’, in particular to check only and with snapshots excluded. Those are most important in Recovery mode, where there can be no alternative.