When the same operation is split across several trace files, analyzing each one alone loses the cursors opened in the others and counts twice the time in which the session was writing to another file. The collection merges the files into a single trace, and the collection card shows how they went in.
What the section shows
When you upload several files with Analyze together, as a single collection (analyze together), they become a single analysis, marked as collection in the History. In the report Summary, right below Session and environment, the collection card appears with one of these titles:
- Collection: N files, analyzed as a single trace, when all the files come from the same session;
- Collection: N files from N sessions, analyzed as a single trace, when there is more than one session.
The card text sums up how the trace was assembled: The segments of each file were merged in time order (tim): a cursor opened in one file is recognized in another, and the interval when the session was writing to another file is not counted twice. With more than one session, it adds: Sessions are not merged: the collection's total time is the sum of each session's time.
The files are cut at the points where Oracle switches files (*** TRACE CONTINUES IN FILE and *** TRACE CONTINUED FROM FILE), and the segments of the same instance, process and session go in in tim order. Different sessions go in one after the other, each with its own cursors. In a collection, the Session and environment card does not show the File on server.
The synthetic trace of the example is a single file, and the sample report does not have this card.
How to read it
The card has two lists that open with a click.
| Column | What it says |
|---|---|
| Session | The session number in the collection, used in the file list. |
| Instance | The database instance where the session ran. |
| Process | The server process of the session. |
| SID.serial | The SID and serial of the session. |
| Time (s) | The response time of the session. The column adds up to the Total time in the Summary. |
| Column | What it says |
|---|---|
| Session | The session of the file. Shown only when there is more than one session. |
| File | The name of the uploaded file. |
| Segments | How many segments of the file went in, in time order. More than one segment means that the session left the file and came back to it later. |
| Rows | How many lines of the file the analysis read. |
A repeated file, or one that is not a trace, is left out, and a warning at the top of the report names it: ... does not look like an Oracle trace and was left out of the collection or ... is identical to ... and was left out of the collection. Segments of different files that overlap in time also raise a warning, so you can check that they belong to the same collection.
In a collection, the Line column of Longest calls and of Errors is the line of the assembled trace, not the line of the original file.
When it is a concern
- The unaccounted-for time between calls is high: a file of the collection may be missing, and the time in which the session was writing to it became a gap. The card for that component in the profile points here (gaps between calls).
- There are more sessions than you expected: files from another operation went in. The sessions are not merged, but the total time adds them all up.
- There is a warning about a file left out or about overlapping segments: check the list before drawing conclusions from the total time.
Where to go next
- Analyze together, to upload the files as a collection.
- Report summary, where the collection card is.
- Response time profile, which adds up the time of all the sessions in the collection.