JENNIFER APM

The dashboards show the JENNIFER console's real-time dashboards inside Sonar. You watch the state of your systems from the Sonar menu without logging in to the console separately. The screens look and work the same way as in the JENNIFER 5.7.0.1 console.

There are four dashboards.

Common layout

All four screens are, from the top: the title bar, the domain bar and the chart cards.

Title bar

On the left is the name of the domain group in view (for example 미분류 도메인, "unclassified domain"); the screen name is in the middle. A screen shows the first domain group the profile's console account can see.

Domain bar

The line under the title bar lists what is monitored, as chips. System Admin, Memory and System Resource show instance chips; Application Insights shows domain chips.

Hover over a chip to see its SID, instance ID with the agent number, and host name. Stopped and unlicensed instances are faded. The selection is remembered within the same browser tab, so it survives moving to another screen and back.

On the right of the domain bar:

• Count - the number of chips. When the chips overflow one line a ▾ appears; click it to unfold the rest
• Scoreboard - running (Live), stopped (Stopped) and unlicensed (UnLicensed) instances
• ⋮ menu - turn the scoreboard on or off (스코어보드 보기), and choose between showing instances of every status (모든 상태의 인스턴스 보기) and hiding stopped ones (정지된 상태의 인스턴스 보지 않기)

Chart cards

Depending on the chart, these icons follow the card title on the right.

Click an instance inside a chart (a line, a bar, a chip) and it becomes selected and is highlighted in the other charts too. Click it again to clear the selection.

Refreshing

There is no refresh interval to set. Every chart fetches new data on its own schedule. Roughly:

Reading the charts

Speed bar

The band across the top of the screen. It animates requests coming in from the left, passing through the cylinder in the middle (transactions running now) and leaving on the right as they are served.

• The number on the left is the arrival rate (requests arriving per second); the number on the right is the service rate
• A number left of the cylinder is the count of rejected requests
• The color of a leaving stream is its response time band

With one instance selected, it shows that instance alone.

Real-time lines

TPS, concurrent users, response time, heap memory and so on, drawn as lines over the last few minutes. Each instance has its own line, and the highest point is labeled. Some metrics can be merged into one line with the card icon.

The response time and concurrent user charts of Application Insights draw the server's confidence band under the line and fill the parts where the line leaves the band in orange. Look there for behavior that is out of the ordinary. Without band values, only the line is drawn.

Today compared

오늘 TPS (today's TPS) and 오늘 동시사용자 (today's concurrent users) draw today's and yesterday's values on one 0-24 h axis. The vertical marker is the current time. Hover to read the value at any time.

Hourly columns

시간당 호출 건수 (calls per hour), 시간당 방문자 (visitors per hour) and 시간당 GC Activity are one column per hour from 0 to 24 h. The current hour's column is highlighted.

Equalizers

The 액티브서비스 (active services) card stands a bar per instance and stacks the transactions running now by elapsed time band, one color per band. The bands are in the legend by the title and follow the console's setting (by default 0-3 s, 3-5 s, 5-8 s, over 8 s). A growing red band means slow transactions are piling up. The card icon switches between a domain view (도메인 단위로 보기) and an instance view (인스턴스 단위로 보기).

The 액티브서비스 현황 (active service status) card shows the same as a cylinder per instance. The number under a cylinder is the count of transactions running in it.

The 시스템 CPU 사용률 (system CPU usage) equalizer on System Resource stacks CPU time as wait, nice, user and sys, with the total above each bar. Its axis is always 0-100. An instance with no values received yet is a grey box.

Events

The EVENT card draws a chip per instance, colored by that instance's most severe event. The counts per level follow the title. The card icon switches between chips (아이콘으로 보기) and a list of the latest events, one per row (리스트로 보기).

With an instance selected (by clicking its chip, for example) that has events, a recovery icon appears after the counts. Read the caution below.

Active applications

The bubble chart of Application Insights. Each application running now is a bubble.

• Across is the average response time, up is the call count
• The bubble size is the slow call count. Hover to see the call count, slow call count and average response time
• The dashed lines are the active service band limits (3, 5 and 8 seconds by default)
• The number by the title is the active service count

A large bubble far to the right is the application keeping users waiting right now.

Application insight

A heatmap of the problems the console server found in the last ten minutes: a row per problem, a cell per minute. With nothing found, the card shows the "no problem detected" (현재 검출된 문제 항목이 없습니다.) graphic.

Stack traces

A heatmap of today's stack trace samples: a column per ten minutes, a cell per minute in it (minute 0 at the bottom), darker for more samples. It is drawn for the day the screen was opened, so after midnight reload the screen to move to the new day. The detailed analysis is Stacktrace Flight Recorder in Analysis.

Application map

A treemap of the applications called in the last few minutes. A tile's size is its call count, and its color is the active service band the chosen metric falls in. The list icon in the card header picks that metric: average response time, average SQL time, average external call time or response time standard deviation. Hover over a tile for its call count, error count and times.

Memory charts

The charts of the Memory screen show the one selected instance.

• 영역별 힙메모리 사용량 (MB) (heap usage by area) has a line per heap area; the switches by the title turn areas on and off
• 영역별 GC 시간 (GC time by area) marks each area's time as a point when a GC happens. Without GCs it is empty
• 시간당 GC Activity (%) is a column per hour of today

GC button

The 힙메모리 사용량 (MB) (heap usage) card shows a garbage collection (가비지 콜렉션) button when the console account may run GC. The button opens the list of the chart's instances; picking one sends the request to run GC on that instance straight to the console. There is no confirmation; the chart shows <instance> - 가비지 콜렉션 for two seconds.

The screens

System Admin

When the selected instances (or all of them, with none selected) reach the count in the console user's settings (10 by default), the layout switches to a wide one and some cards, such as active service status, are hidden.

Application Insights

Seen by domain. It has the speed bar, active applications, TPS, application insight, response time (with confidence band), concurrent users (with confidence band), stack traces, application map, X-View and calls per hour. Look for symptoms in application insight and active applications, and for candidate causes in the application map and the stack traces.

Memory

One instance at a time: heap usage by area, GC time by area, process CPU usage, heap usage (with the GC button), GC time, system CPU usage, non-heap usage, GC activity, system memory usage, object pending finalization count, GC activity per hour and process memory usage.

Heap usage that rises and drops in a sawtooth is a normal GC cycle. A floor that keeps rising, or GC activity that stays high, points to a memory leak or a heap that is too small.

System Resource

Several instances side by side: the system CPU usage equalizer, the process CPU usage equalizer, and real-time lines of system CPU usage, system memory usage (%), process CPU usage, system memory usage (MB), process memory usage and heap usage (with the GC button).

The X-View chart

The X-View on System Admin and Application Insights puts every finished transaction on the chart as a point: across is the time it ended, up is its response time. Transactions that ended in error are drawn in reds, on top of the normal ones. Select an instance in the domain bar or another chart and only its points stay.

Header

The filter switches combine with AND.

• ERROR, MSA 시작점 (MSA root), 스택트레이스 (stack trace), 비동기 (asynchronous), 동기 (synchronous) - only transactions of that kind. Asynchronous and synchronous cannot be on together
• 대표 트랜잭션 (representative transactions) - only when the console has AI turned on. Keeps the transactions that represent the distribution; it cannot be combined with the other switches
• ≤ P90, ≤ P95, ≤ P99 - only points at or below that percentile of the vertical axis

The Y axis menu picks the vertical axis among response time, SQL time, fetch time, external call time and CPU time, and 자동 Max 값 (automatic max) makes the maximum follow the points.

The ⋯ menu has the X점 크게 보기 (large X marks) and Deploy 이력 보기 (deploy history, with the permission to show deploys) switches. Deploy history puts a marker at each deploy.

Pattern badge

With AI and X-View pattern recognition on in the console, a badge follows the title. Every few seconds it judges the distribution of points and shows normal (정상 상태), abnormal (비정상 상태) or the name of a registered pattern, with its likelihood (for example 정상 상태 일 가능성이 100%입니다., "100% likely to be normal"). Hover for the pattern's description and the sensitivity setting (automatic or manual). Without any pattern registered in JENNIFER, no badge shows.

Keyboard

Click the chart once to give it the focus first.

Dragging

Drag a box over the points: the count of points in the box shows while you drag, and on release those points open in the X-View Transactions window. A box with no points opens no window. What you can do in that window is in X-View Transactions.

Actions that change the console

The dashboards mostly read, but the following two change the state of the JENNIFER server with the profile's console account. Both appear when the account has the permission, and run at once without confirmation.