How it works
One agent, every host.
01 · Install
One agent
A single errorgap-agent per host, with AI-assisted setup that reads your OS and writes the install command.
02 · Collect
Snapshots every 15 s
CPU, memory, disk, processes, containers, interfaces and mounts.
03 · Share
One host, several projects
A shared host reports once and appears in each project that uses it.
04 · Correlate
Pressure becomes evidence
Saturation on a host shows up on the incidents it overlaps with.
What you see
From the fleet down to the mount.
Fleet and host pages
Sort the fleet, open the host.
Sortable hosts, volumes and processes. Load average is explained against core count, so 3.6 on four cores reads differently from 3.6 on one.
- Hosts, volumes and processes views
- Load average explained per host
- Live updates without refresh
Hosts · 6sorted by CPU ▾
HostCPULoad
worker-184%3.6
web-147%1.8
web-239%1.4
db-122%0.9
Containers
Docker containers, side by side.
CPU and memory per container on each host, so the noisy one is obvious.
- CPU and memory per container
- Uptime per container
Containers · web-1docker · 5 running
ContainerCPUUp
app31%6d 4h
sidekiq18%6d 4h
nginx3%6d 4h
cache2%6d 4h
errorgap-agent1%6d 4h
Network and volumes
Interfaces and mount options.
Throughput per interface, and every mount with its options, so a missing noexec on /tmp is one glance away.
- rx / tx per interface
- Mount options and usage per volume
Network & volumes · db-1
Ifacerxtx
eth04.2 MB/s1.1 MB/s
eth1820 KB/s760 KB/s
MountUsed
/53%
/var/lib/pg71%
/tmp4%
Time ranges and compare
Compare with the previous period.
Pick any range, overlay the period before it, and hover to read both. Hover behaves the same as on the Performance charts.
- Custom time ranges
- Previous-period overlay on every chart
- Hover matched across Infrastructure and Performance
CPU · worker-1Oct 6 09:00 – 15:00
11:15 · 84%
previous · 45%
this periodprevious period
Related capabilities