Layered
Hardware, operating system, services and databases each have their own tool and their own owner, and nobody traces across them.
Solution · By Stack
Servers & Databases
When an application slows down, the cause can be the hardware, the operating system, a stopped service, or a database under pressure. Unified360 watches every layer from the iDRAC to the wait class, for Windows and Linux servers and four database engines, so you can trace a symptom down the stack in one place.
- ServersWindows and Linux
- Databasesfour engines
- HardwareiDRAC and iLO
Trace a symptom down the stackPick what users are seeing · sample data
- HardwareDell iDRAC and HPE iLO: health, temperature, power, fans and disks
- Operating systemWindows and Linux: CPU, memory, disk by volume, network and processes
- Services and filesService state, file freshness and port connectivity
- Database engineSQL Server, Oracle, PostgreSQL and MySQL: availability, connections, cache, replication
- Queries and waitsLocks, deadlocks, wait classes, slow queries and transaction rates
The problem
Trouble hides between the layers
Each team sees its own layer. The symptom lives in all of them.
Reactive
Disks fill, services stop and connections run out. The first alert is a phone call from a user.
Opaque
The instance is up, but nobody can see the waits, the locks or the replication lag that explain why it's slow.
Where it goes wrong
Six gaps, and how each one is closed
Servers, databases and hardware in three tools
How it goes wrongThe server team, the DBAs and the hardware team each watch a different console.
Closed withOne platform covers Windows and Linux servers, Dell and HPE hardware controllers, and four database engines, with a Server Overview and a Database Overview across the fleet.
Disks that fill without warning
How it goes wrongA volume creeps toward full and the first alert is the outage.
Closed withDisk usage and free space are evaluated per volume with separate state for each mount, and Oracle tablespaces get their own alert per tablespace. SQL Server shows volume space too.
Services that quietly stop
How it goes wrongThe server is up and pinging, but the service that matters isn't running.
Closed withService state rules, a Service Status table on each server, port connectivity checks, and a file monitor that alerts when a watched file hasn't changed for a set time.
Databases that are up but unhealthy
How it goes wrongThe instance answers a ping while connections, caches and locks head for trouble.
Closed withConnection use, cache hit rate, deadlocks, replication lag, tablespace use, page life expectancy and wait classes, each with warning and critical thresholds.
Hardware failures nobody sees
How it goes wrongA fan, a power supply or a disk degrades and nobody looks at the controller.
Closed withDell iDRAC and HPE iLO are polled for health, power and temperature, and iDRAC dashboards add fans, physical disks, RAM slots and the system log.
Fixing it by hand, every time
How it goes wrongThe same cleanup or restart, typed by whoever gets paged.
Closed withAutomated Actions can run approved scripts on Windows and Linux servers when an alert fires, with cooldowns, dry runs and approval for destructive steps.
Servers
Everything about a server, on one page
CPU, memory, disk by volume, disk I/O, network, processes, and the state of every service, for Windows and Linux.
Below the charts, the same page lists processes and threads, disk read, write and I/O, the state of each service with a filter, file monitoring, port connectivity and, on Windows, exception dispatches.
Databases
Four engines, one way of watching them
Pick an engine to see the metrics Unified360 tracks and the sections of its dashboard.
SQL Server
Instance health, memory, batch and transaction rates, and the counters that show pressure.
- cpu_pctSQL CPU
- memory_usage_pctMemory in use
- buffer_cache_hit_ratioCache efficiency
- page_life_expectancySeconds a page stays in memory
- batch_requests_secBatch requests per second
- lock_waits_secLock waits per second
- deadlocks_secDeadlocks per second
- sql_compilations_secCompilations per second
- db_offline / db_suspectDatabases needing attention
- uptime_hoursInstance uptime
Runs on-premises, and covers Azure SQL Database, Azure SQL Managed Instance and Azure SQL Pool.
Oracle
Availability, sessions, commits, and the wait classes that show where time goes.
- db_statusUp or down
- active_sessionsActive user sessions
- tablespace_usage_pctUsed % per tablespace
- process_countProcesses
- user_commits / user_rollbacksTransaction outcomes
- execute_countSQL executions
- wait_concurrencyConcurrency waits
- wait_commitCommit waits
- wait_user_ioUser I/O waits
- wait_system_ioSystem I/O waits
A tablespace rule can watch every tablespace at once, each with its own alert state. Wait times are cumulative counters.
On the dashboard
Sections of the Oracle detail page.
Instance SnapshotTablespacesPeak Tablespace UsageResource UtilizationWait Time SummaryWait TrendIndex Usage & Storage
PostgreSQL
Connections, cache, deadlocks, replication and transaction activity.
- db_statusUp or down
- active_sessionsActive sessions
- connection_pctShare of max connections
- cache_hit_rateBuffer cache hit rate
- deadlocks_per_secDeadlocks per second
- replication_lag_secReplication lag in seconds
- idle_in_transactionIdle-in-transaction sessions
- total_locksActive locks
- wal_size_bytesWAL size
- archiver_failedWAL archiver failures
Databases on the server are discovered automatically.
On the dashboard
Sections of the PostgreSQL detail page.
Server SnapshotSession ActivityTransactions/secTuple Operations/secDatabase SizesServer Settings
MySQL
Threads, connections, the InnoDB buffer pool and slow queries.
- db_statusUp or down
- threads_connectedConnected threads
- connection_pctShare of max connections
- slow_queries_per_secSlow queries per second
- innodb_row_lock_waitsInnoDB row lock waits
- buffer_pool_used_pctBuffer pool used
- aborted_connectionsAborted connections
- threads_runningRunning threads
- qpsQueries per second
- tmp_disk_tablesTemp tables on disk
Connection use and slow-query rate come with warning and critical thresholds.
On the dashboard
Sections of the MySQL detail page.
Server SnapshotConnection ActivityDML Commands/secInnoDB Buffer PoolNetwork I/OServer Settings
Metric names are the ones you can build alert rules on, each with warning and critical thresholds. Only the SQL Server screen is shown here.
What changes
When every layer is in view
One view from hardware to query
Servers, databases and controllers report into the same platform, so a symptom can be traced down the stack.
Space and services under watch
Volumes, tablespaces, services and files each have rules, with recovery when they clear.
Database pressure you can see
Connections, cache, locks, deadlocks, waits and replication lag are visible before users complain.
Fixes that start themselves
Approved scripts can run when an alert fires, with the guardrails of Automated Actions.
Built with
The platforms behind this solution
Unified360
Operations and observability
Server, database and hardware monitoring, fleet overviews, alerting, and Automated Actions for approved fixes.
IntelliAsset
Asset and patch management
Hardware and software inventory for your servers, and patch scanning and rollout for Windows and Debian and Ubuntu.
Questions
Frequently asked
Which operating systems and databases are covered?
Windows and Linux servers, through the Grafana Alloy agent, and SQL Server (including Azure SQL Database, Managed Instance and Pool), Oracle, PostgreSQL and MySQL. Tomcat, JBoss and IIS have their own application monitors.
Does it show individual slow queries and plans?
No. It tracks instance counters and rates, such as batch requests, lock waits, deadlocks, wait classes and slow-query rate, and alerts when they cross a threshold. It isn't a query profiler.
How is server hardware monitored?
Dell iDRAC and HPE iLO controllers are polled over SNMP v2c for health, power and temperature. iDRAC dashboards also show fans, physical disks, RAM slots and system log entries, and the controller reports even when the operating system doesn't.
How do disk alerts work?
Rules evaluate disk use and free space per volume, keep separate state for each mount, fire after a set number of consecutive checks and recover when the volume drops back. Oracle tablespaces get their own alert per tablespace.
Can it tell me when a service or a file stops?
Yes. Windows service state rules alert when a service isn't running, port connectivity checks confirm a service is reachable, and a file monitor alerts when a watched file hasn't been modified for a set number of seconds.
Can it fix problems automatically?
Yes, through Automated Actions. An alert can start an approved script on a Windows or Linux server, with a cooldown, a dry-run mode and an approval for destructive scripts.
Do I need something installed on every server?
Yes for operating-system metrics: the Alloy agent collects CPU, memory, disk and network and forwards them to Unified360. Databases and hardware controllers are reached from a monitoring agent at the site.
Can I see the whole fleet at once?
Yes. The Server Overview ranks servers by CPU, memory, disk and network and shows the Windows and Linux split, and the Database Overview shows engine mix, connection use, cache hit rate and replication lag.
See it on your own servers
Trace your next slowdown from the query to the fan
Tell us which servers and databases you run and we'll show you Unified360 watching them. No commitment required.