Sending server monitoring.
Disk, CPU
and queue.

Four fleet status tiles: how many sending servers are healthy, need attention, have failed or have gone silent. This is the first look at server monitoring — is everything fine with the hardware and the MTA. In the example, sixteen servers of seventeen are healthy and one needs attention — mq-01, with its disk 81% full; there are no failures and no silent servers. Clicking a tile leaves only the servers in that status in the table below. The Hub takes measurements from the servers itself, once a minute. Read the tiles by urgency: failure — the MTA isn't responding or the disk is 90% full or more; attention — disk at 75–90% or CPU at 85% or more; silent — no measurement for more than 15 minutes.

Enlarge

• Healthy — 16 of 17 — the MTA is running, disk and CPU within limits.
• Attention — 1 — mq-01, disk 81%.
• Failure and silent — none.

The tiles work as a filter for the table below.

The tab's main table: all 17 sending servers, one row each — status, score, MTA, queue, CPU, memory, disk, messages sent in the last hour and the time of measurement. Use it to compare servers with each other and spot which ones are different. In the example, mq-01 is in “attention” because of its disk at 81%, mq-02, mta-06, mta-07 and mta-10 have a lower score, and lab-01 and mta-11 haven't sent anything yet. The “Measured” column shows how fresh the numbers in the row are. mq-02 scores 84 because of 217 bounces for mailbox provider requirements: that's the address Microsoft rejects, so the hardware isn't the issue. Only the disk can turn red in the bars.

Enlarge

• mq-01 — “attention”: disk 81%, score 89.
• MailerQ — 145–152 in the queue, 41–43 an hour.
• tx-01 and tx-02 — 39–66 in the queue, 54–62 an hour.
• mq-02 and mta-10 — scores 84 and 80.
• lab-01 and mta-11 — “not enough data” — nothing was sent.
• Measured — how fresh the numbers are.

The bars are colored by the status thresholds: only the disk can turn red.

The table after clicking the “Attention” tile: one server is left — mq-01, and the caption says “showing 1 of 17”. This filter helps when you have many servers but only need to look at the problem ones. The row shows right away what's wrong: the disk is 81% full. The “attention ×” button above the table clears the filter and brings back the whole fleet. The “Failure” and “Silent” tiles work the same way: failed and silent servers are always shown first. The table shows working servers only; those being prepared or retired are shown in “Infrastructure”, and the table caption gives their number. Load is still compared with servers on the same MTA.

Enlarge

• Showing — 1 of 17 — mq-01.
• Clear the filter — the “attention ×” button above the table.

This way one problem server doesn't get lost among sixteen healthy ones.

The “Who stands out from the fleet” block lists servers that differ from the rest, each with its reason. In the example there are seven of 17: mq-01's disk is 81% full and it's worth checking what's growing — the logs or the queue; mta-06, mta-07, mta-10 and mq-02 have a dropped score; mta-10 and mta-11 have addresses without a reverse DNS record. Gray at the bottom are lab-01 and mta-11, which haven't sent any reports yet because they haven't sent any messages either. Work through the rows from top to bottom. The mq-01 score, dropped because of the same disk, doesn't get a second row: one problem — one row. A single server in full — MTA status, queues and controls — is examined in monitoring.

Enlarge

• mq-01 — disk 81% — logs or queue.
• mta-10 — score 80 — spam and complaints.
• mq-02 — score 84 — mailbox provider requirements.
• mta-10 and mta-11 — one address without PTR each.
• In gray — lab-01 and mta-11 — no reports, and no messages either.

One problem — one row.