Ceph » mgr » Dashboard

Description of problem¶

The Read Bytes, Write Bytes, Read ops, and Write ops values freeze after an OSD is down.

How reproducible¶

• Generate some I/Os to a pool.
• Shutdown a host.
• Observe the I/O stats on OSD list table.

Actual results¶

Screenshot after killing host node3 (I/Os are ongoing)
¶

Screenshot after killing host node3 (I/O are canceled and everything is settled down)¶

Expected results¶

The dot-graph or Read-Write ops should become 0.

Additional info¶

Here are stats_history data return from backend:

DOWN OSD

"stats_history":{

    "op_w":[
        [
            1607575596.8253326,
            0.0
        ],
        [
            1607575601.8246965,
            0.0
        ],
        [
            1607575606.8268645,
            0.0
        ],
        [
            1607575611.8271983,
            0.0
        ],

IN/UP OSD

"stats_history":{

    "op_w":[
        [
            1607576260.8144872,
            0.0
        ],
        [
            1607576265.8149185,
            0.0
        ],
        [
            1607576270.815149,
            0.0
        ],
        [
            1607576275.8155153,
            0.0
        ],

The timestamps in the DOWN OSD freeze at the moment that the OSD is down.
The timestamps in the UP OSD advance with time.