I’ve recently encountered numerous failures of the eMMC drives on two of my Mixtile Blade 3 SBCs (four in a Ceph cluster): The drive announces a ”cache flush error” (errno 110), goes offline, and is then remounted in read-only state:
[93724.175631] mmc0: cache flush error -110
[93724.175689] I/O error, dev mmcblk0, sector 4584968 op 0x1:(WRITE) flags 0x29800 phys_seg 1 prio class 2
[93724.175731] Aborting journal on device mmcblk0p3-8.
[93724.177134] EXT4-fs error (device mmcblk0p3): ext4_journal_check_start:83: comm systemd-journal: Detected aborted journal
[93724.177227] EXT4-fs error (device mmcblk0p3): ext4_journal_check_start:83: comm rs:main Q:Reg: Detected aborted journal
[93724.177572] EXT4-fs (mmcblk0p3): Remounting filesystem read-only
Of course, I can’t use a cluster node, which has its system drive (O/S, config, logs, containers etc.; the Ceph datastore itself lies on a separate SSD) set to read-only.
The O/S is Ubuntu 24.04 LTS for ARM64. The eMMC drive runs in HS200 mode without external strobe (HS400 led to a bunch of system freezes). I’ve already told the manufacturer about the issue, but haven’t received a response so far. What’S wrong here?
