Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

NTFS event 55 and 98: 'The file system structure on the disk is corrupt and unusable' on Windows Server

Technical article
Troubleshooting

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

Event 55 reads "The file system structure on the disk is corrupt and unusable. Please run the chkdsk utility on the volume." Event 98 says the volume "needs to be taken offline to perform a Full Chkdsk". Event 55 reports a damaged structure and asks for chkdsk; event 98 says a full offline chkdsk is needed.

Take a backup of the volume before anything else, then scan, read the storage events around the same time, and schedule the repair. Microsoft's guidance is to check the hardware path as well as run chkdsk, because Microsoft's Engage Center page ties event 55 to a failure condition with the disk subsystem.

What the error means

Microsoft's data corruption troubleshooting page lists event 55 as The file system structure on the disk is corrupt and unusable. Please run the chkdsk utility on the volume. and event 98 as Volume C: (\Device\HarddiskVolume3) needs to be taken offline to perform a Full Chkdsk. Please run "CHKDSK /F" locally via the command line or run "REPAIR-VOLUME \<drive:>" locally or remotely via PowerShell. The drive letter and device number in your own event will differ.

The same page groups events 55, 50, 140 and 98 together as NTFS events that "indicate file system corruption and similar issues". It gives two conditions behind such corruption: a disk with bad sectors, or I/O requests from the file system that the disk subsystem did not complete successfully. It adds that NTFS may be unable to write to its transaction log, which can affect its ability to stop or roll back the operations whose log data was not written.

Microsoft's Engage Center page on event 55 says the event is "typically related to a failure condition with the disk subsystem and may stop the machine from functioning correctly". It also says that if there is any indication of a structural problem, the disk should be replaced as soon as possible, and that for a SAN connection (LUN) you should investigate access to the SAN.

Event 98 is a request, not a result: Windows is saying an online scan or fix is not enough and a full offline check is needed. Resilient File System (ReFS) volumes are different. The same Microsoft page notes that "Resilient file system (ReFS) volumes can automatically fix corruption issues", and that chkdsk can still scan them.

Likely causes

Cause What you see How to confirm
Bad sectors on a physical or virtual disk Event 55 or 98 with other disk errors; chkdsk reports bad sectors chkdsk /scan output, plus a read of the storage events listed in Microsoft's checklist
Incomplete I/O between the file system and the disk subsystem Event 55 or 98 near event 129, 153 or 157 in the same minutes System log timeline across all five event IDs
Controller, driver or firmware fault Errors across several volumes on one controller or path Driver and firmware versions against the hardware vendor's current release
Filter driver interference Errors persist after a clean chkdsk Microsoft's advanced steps: remove or update filter drivers, remove third-party disk-management software
SAN or multipath fault Errors on a LUN only, with path changes logged SAN access and MPIO configuration, as Microsoft's page directs
Interrupted write (the cause is advice, not a Microsoft statement) Volume reports dirty; event follows a crash fsutil dirty query result

Events 129 and 153 are storage timeouts; Microsoft's page describes 129 as logged when the Storport driver times out a request to the disk, and 153 as logged when the Storport miniport driver times out a request. Event 157 is a surprise removal of a nonremovable disk. If you see them near event 55, the cause sits below the file system.

Prove the cause before changing anything

Run these in an elevated Command Prompt or PowerShell. They read; none of them is documented as a repair.

  1. Open the System log and list events 55, 98, 50, 140, 129, 153 and 157 for the past few days. Note the first time any of them appears, as a starting point for the timeline.

  2. Get volume details:

    fsutil fsinfo ntfsinfo X:
    
  3. Check whether the volume is marked dirty:

    fsutil dirty query X:
    

    Microsoft's page says that if the result shows the volume is dirty, you should schedule a maintenance window, because the disk is not accessible during repair.

  4. Scan without choosing a repair:

    chkdsk /scan
    

    The troubleshooting page describes this as scan mode that looks for issues "without making changes". The chkdsk reference says /scan "runs an online scan on the volume" and documents /forceofflinefix as the switch that bypasses all online repair, which implies an online scan may attempt an online fix. Treat /scan as low risk, not as guaranteed change-free, and take the backup before you run it on a damaged volume.

    The PowerShell equivalent is Repair-Volume -DriveLetter X -Scan. The Repair-Volume page says -Scan scans the volume "without attempting to repair it" and adds detected corruptions to the $corrupt system file.

  5. Read earlier chkdsk and wininit results from the Application log:

    get-winevent -FilterHashTable @{logname="Application"} | ?{$_.providername -match "chkdsk"} | fl timecreated, message
    

    Run the same with wininit in place of chkdsk to read the boot-time check result.

How to read the results: a scan that finds no errors, with event 55 or 98 still recurring, points away from the file system and towards the hardware path or a filter driver. A scan that lists corruption, with 129, 153 or 157 earlier in the timeline, points to the disk subsystem. Microsoft's pages give no threshold for deciding between them, so take the earliest event in the timeline as the place to start.

Fix

Every fix starts with a backup. If event 98 says the volume needs offline repair, a backup of that volume may itself fail or read damaged data, so check the backup finished cleanly. Microsoft's prerequisites say "Before you perform any repairs or modify any disks, back up all critical data." The page also says that if chkdsk does not fix the errors, "you have to restore the data from a backup."

Cause: dirty or damaged NTFS structure (events 55, 98).

Disruptive step: chkdsk /f /r locks and repairs the volume, and the volume is unavailable while it runs. Microsoft says /f "fixes errors on the disk" and requires the disk to be locked; /r "locates bad sectors and recovers readable information" and includes /f. If chkdsk cannot lock the drive it offers to run at the next restart. There is no way back from a completed repair other than restoring from your backup. Microsoft notes that interrupting chkdsk is not recommended, though interrupting it "shouldn't leave the volume any more corrupt than it was before".

chkdsk /f /r

On a large volume or one with millions of files, Microsoft warns chkdsk /f can take a long time. Schedule enough time and tell the people who use the share.

Cause: you want less downtime.

Disruptive step: Repair-Volume -DriveLetter X -SpotFix (or chkdsk /spotfix) takes the volume briefly offline and fixes only issues already logged in the $corrupt file by a prior scan. Back out by restoring from backup; a completed fix cannot be undone.

Repair-Volume -DriveLetter X -SpotFix

Repair-Volume -DriveLetter X -OfflineScanAndFix takes the volume offline to scan fully and fix, and is documented as equivalent to chkdsk /f.

Disruptive step: -OfflineScanAndFix makes the volume unavailable for the whole scan and repair, and a completed repair cannot be undone. Back out only by restoring from your backup.

Cause: open handles stop chkdsk locking the volume. Microsoft says chkdsk cannot correct errors while files are open and offers to run at restart. /x forces a dismount.

Disruptive step: chkdsk /x forces the volume to dismount and, in Microsoft's words, "All open handles to the drive are invalidated." Applications with open files lose them; unsaved data in those files can be lost. There is no way back for a handle already invalidated.

Cause: hardware, driver or filter-driver fault. Microsoft's list: update storage drivers and firmware (iSCSI port, RAID controller, HBA, DSM, MPIO), remove or update filter drivers, uninstall third-party disk-management software such as Diskeeper, and test with a different driver type. For event 55, the Engage Center page says to replace the disk if there is any indication of a structural problem. Risk: replacing a disk or changing a storage driver is a hardware change; plan a rebuild or restore path first.

Example (illustrative): fs01 hosts a 4 TB volume D: holding 1.2 TB of department files. At 02:10 the System log shows event 153 for Disk 2, at 02:12 event 55 on D:, and at 07:30 event 98 appears when a user opens a share. The administrator copies the 1.2 TB (1,200,000 MB) to a second server over a link that sustains 200 MB/s, which takes 1,200,000 / 200 = 6,000 seconds, or 100 minutes. After the copy, fsutil dirty query D: reports the volume dirty, and a maintenance window is booked for chkdsk /f /r. The retry event 153 before event 55 tells the administrator to ask the storage team about the path after the repair.

Confirm the fix

  • Run chkdsk /scan again; the chkdsk reference lists exit code 0 as "No errors were found" and 1 as "Errors were found and fixed".
  • Run fsutil dirty query X: and confirm the volume is no longer dirty.
  • Read the chkdsk and wininit entries in the Application log to see what the repair changed.
  • Confirm that events 55 and 98 have stopped across at least one full backup cycle and the busiest period of the week.

What to watch afterwards

  • Watch for events 50, 55, 98, 129, 140, 153 and 157 in the System log. A return of any of them after a clean repair means the cause was not the file system.
  • Record the disk, controller and firmware versions you updated, so a recurrence can be matched to a change.
  • Keep a restore test for this volume in the schedule. Microsoft's guidance is that persistent errors may need a restore from backup.
  • On solid-state disks, Microsoft cautions that repeated full-surface scans with /r can add write and erase cycles, which "slightly" reduce lifespan. Use them when needed, not as routine.

Risks and limits

These commands have not been run or lab-tested. Event text comes from Microsoft Learn pages and your own event will show your drive and device numbers. Microsoft's pages give no repair time estimates and no threshold for replacing a disk beyond "any indication of a structural problem".

Running chkdsk against a source volume while Shadow Copies for Shared Folders is enabled can report false errors or quit unexpectedly, because the volume cannot be locked. Microsoft's suggestion for checking shadow copies is to run chkdsk in read-only mode (without parameters) against the Shadow Copies for Shared Folders storage volume. A repair changes the file system, not the underlying hardware, so a failing disk will produce the same events again.

Sources