#95 Implement cache purging (admin + non-admin) with timing + logging #97

vpstackhub · 2025-12-05T21:07:11Z

Summary

This PR implements the cache purge feature requested in Issue #95, providing both:

Admin-level purge via EmptyStandbyList.exe

Fallback non-admin purge (PowerShell write-flush trick)

Detailed timing logs for James to review the UX impact

Clear console debug messages to distinguish pathways

No GUI changes yet. Waiting for James' preference on how to present timing (progress bar, spinner, modal, etc.)

Test Results

TEST A -- Running inside NetBeans (non-admin)

BenchmarkWorker: ENTERING CACHE PURGE SECTION
endTime=.... (~50ms)
startTime=.... (~30ms)

**TEST B **--Running packaged build with Administrator privileges

BenchmarkWorker: ADMIN purge path selected
BenchmarkWorker: Calling UtilOs.dropOsCache()
DROP CACHE CALLED inside UtilOs
dropOsCache() returned = true
Total purge time ~1.0s

Both test paths behaved exactly as designed.

Consideration-
Based on the timing results above (~1 second for admin purge, ~80ms for fallback),
I'd love your recommendation for the GUI implementation:

1.simple toast message?

2.small modal?

3.mini-progress bar?

4.indeterminate spinner?

Happy to adjust based on your preference.

src/jdiskmark/BenchmarkWorker.java

jamesmarkchan · 2025-12-07T21:12:24Z

src/jdiskmark/RunDetailsPanel.java

nice this appears to be a pure swing panel, no generated code! awesome :)

src/jdiskmark/Util.java

jamesmarkchan · 2025-12-07T21:21:35Z

src/jdiskmark/UtilOs.java

+        }
+    }
+
+    public static void restartAsAdmin() {


may want to indicate this is only windows since i see powershell syntax here, should warn that generally i don't think it's possible to do privilege escalation but if there are tricks that allow us to escalate easily that would be good knowledge for all of us. I think the normal approach is we check if it is an admin process and if not we don't attempt to drop the cache and inform the user that they must drop it manually. actually i think if they are not admin we just use your new disk cache purge function right? or is this method used?

jamesmarkchan · 2025-12-07T21:37:20Z

nbproject/project.properties

@vpstackhub what version of netbeans are you using, interesting that this got checked in which seems to break running in netbeans for others?

specifically this line because it does not exist for other users.

project.dir=C:/Users/valer/OneDrive/Documenti/NetBeansProjects/jdm-java

Might you know why these attributes are being added into the project file?

jamesmarkchan · 2025-12-07T21:48:40Z

src/jdiskmark/UtilOs.java

+        }
+    }
+
+    public static boolean dropOsCache() {


@vpstackhub is this the more robust version of Gui.dropCache()? If so it is valid to treat macos and linux the same which Gui.dropCache() does not?

jamesmarkchan · 2025-12-07T22:15:38Z

@vpstackhub Looks really good thanks to your pull request I was able to follow some of the action!

Two key items to look at are:

There is a hard coded path in the project file we'd want to under stand if it is needed or can just be removed.
The purge size might need to be dynamic to be the size of system memory or potentially a percentage of it

I think progress bar and some message notification in the log area is a responsible way to inform the user of the underlying operation. If running in command line mode we can output to the console that we are purging and there are ways to adjust the command line progress bar as well.

Good progress on this objective which will make packaging requirements less painful and more robust for @jslcom @tylermlui and me. <3<3<3 :)

jamesmarkchan · 2025-12-07T21:59:59Z

src/jdiskmark/BenchmarkWorker.java

+
+            // Detect purge size roughly equal to the full test data footprint
+            long estimatedBytes = (long) App.numOfSamples * (long) App.numOfBlocks * 
+                                  ((long) App.blockSizeKb * 1024L);


One thing that we might run into is if the bytes we purge is less than the actual disk cache the benchmark would still be affected by the disk cache. When i asked gemini it recommended purging the size of the entire main memory to be safe as that was the only way to guarantee that the disk cache does not affect our benchmark even if the disk cache is usually just a portion of main memory and for this reason if we go the safe route the the process might be a little lengthy since different system could have different amounts of memory installed, might be worth notifying user what is it doing so they understand why the clearing is taking so long. maybe we can default to purging the size of system memory + 10% (need to detect it from ), i think it's simply Runtime.getRuntime().totalMemory(); and maybe later can add an advanced option for the user to do the cache purging based on a percentage of the main memory if we find safe values for this. @jslcom are you familiar with the default disk cache sizes?

jamesmarkchan · 2025-12-07T22:21:01Z

src/jdiskmark/Util.java

        return (path.delete());
    }

+    public static void readPurge(long estimatedBytes) {


what do you think about purgeDiskCache()?

Is it possible to read the disk cache to achieve purging? No writing is needed? I was under the impression we had to write the amount of the disk cache. Or is the code writing but it just wasn't obvious to me?

jamesmarkchan · 2025-12-15T09:34:23Z

nbproject/project.properties

 src.dir=src
 test.src.dir=test
+run.working.dir=${project.dir}
+work.dir=${project.dir}


@vpstackhub to run the project i needed to remove these lines but I'm not sure if the cache is executing. Possible we can meetup this week so I can show you what I'm seeing? Want to make sure I'm running the branch correctly.

val if you replace these lines does everything still work for you? if so maybe we can remove them or maybe can let me know what these lines will do for us?

jamesmarkchan · 2025-12-15T09:35:13Z

src/jdiskmark/App.java

     */
+    /**
+    * Initialize the GUI Application.
+    */


this is super minor but lines 173-174 lost a valid space.

jamesmarkchan · 2025-12-15T09:38:07Z

src/jdiskmark/Benchmark.java

+        cachePurgePerformed = false;
+        cachePurgeSizeBytes = 0L;
+        cachePurgeDurationMs = 0L;
+        cachePurgeMethod = CachePurgeMethod.NONE;


If interested lines 200-203 and 209-212 can be set at declaration I think since the values are constants.

jamesmarkchan · 2025-12-15T09:43:55Z

src/jdiskmark/BenchmarkWorker.java

    }
+
+    private void printCliProgress(int pct) {
+        int bars = pct / 5;      // 20-bar progress


super minor, there is a java practice where the recommendation is to avoid abbreviations when possible so percent instead of pct.

jamesmarkchan · 2025-12-15T09:46:52Z

src/jdiskmark/UtilOs.java

+            Process p = Runtime.getRuntime().exec("net session");
+            p.waitFor();
+            return p.exitValue() == 0;
+        } catch (Exception ex) {


Generally catching Exception is work in progress / temporary code and RuntimeException less so. might be safe to catch RuntimeException here instead to avoid the appearance of eating checked exception.

Thanks James,appreciate the thorough review. To reduce the back and forth, here’s a consolidated checklist (maybe more helpful for me than it is for you 😂) of what’s done and what I’ll follow up with next so we’re fully aligned.

⸻

Confirmed / Already Addressed
• [x] Removed NetBeans-specific project settings
Deleted run.working.dir and work.dir from nbproject/project.properties.
After removal, the project runs correctly on my side and cache purging still executes as expected. I agree these lines should not be committed.
• [x] Replaced purge method string literals with enum
Introduced Benchmark.CachePurgeMethod and updated all usages (GUI + CLI).
• [x] Resolved local DB enum mapping issue
Cleared .jdm/0.6.3-dev to regenerate schema cleanly --no further persistence errors.
• [x] User feedback during purge
• GUI log messages indicating purge start/end + method used
• CLI progress bar output when running in command line mode

⸻

Minor Cleanup / Follow-ups (next commit)
• [x] Remove duplicate Javadoc block + restore spacing in App.java
• [x] Rename pct → percent in BenchmarkWorker.printCliProgress() for readability
• [x] Narrow catch (Exception) -> catch (RuntimeException) in UtilOs.isProcessElevated()
• [x] (Optional) Move default cachePurge* values to field declarations in Benchmark

⸻

Once the above cleanup is in, I believe PR #95 should be fully ready from my side.

this should make cache behavior much better

jamesmarkchan · 2025-12-28T23:47:25Z

src/jdiskmark/Benchmark.java

    BenchmarkType benchmarkType;
    public BenchmarkType getBenchmarkType() { return benchmarkType; }

+    public enum CachePurgeMethod {


@vpstackhub I was thinking how about calling the top level operation CacheResetMethod this way if we know we need to do a cache reset based on the user being admin or not we can either DROP (admin removal) or PURGE (cache overwrite). This way the parent operation is called something different from the actual mechanism used.

public enum CacheClearanceMethod { NONE, DROP, // Requires Admin PURGE // Fallback for Non-Admin }

alternative could use CacheClearanceMethod or CacheMitigationMethod.

jamesmarkchan · 2025-12-28T23:49:27Z

src/jdiskmark/BenchmarkWorker.java

            Gui.mainFrame.refreshWriteMetrics();
        }
+
+        System.out.println(">>> BenchmarkWorker: ENTERING CACHE PURGE SECTION");


maybe this could be used: ENTERING CACHE RESET SECTION or ENTERING CACHE MITIGATION SECTION

jamesmarkchan · 2025-12-28T23:52:02Z

src/jdiskmark/BenchmarkWorker.java

+
+            long purgeStartNs = System.nanoTime();
+            benchmark.cachePurgePerformed = true;
+            benchmark.cachePurgeMethod = App.isAdmin


potentially cachePurgePerformed > cacheResetPerformed

and cachePurgeMethod > cacheResetMethod

jamesmarkchan · 2025-12-28T23:53:16Z

src/jdiskmark/BenchmarkWorker.java

+            // Detect purge size roughly equal to the full test data footprint
+            long estimatedBytes = (long) App.numOfSamples * (long) App.numOfBlocks * 
+                                  ((long) App.blockSizeKb * 1024L);
+            benchmark.cachePurgeSizeBytes = estimatedBytes;


this is set from zero to the actual purge size and only if not admin and a purge operation occurs.

#95 – Implement cache purge fallback (admin + non-admin) + timing logs

91893a1

jamesmarkchan changed the base branch from main to dev December 7, 2025 20:22

jamesmarkchan reviewed Dec 7, 2025

View reviewed changes

src/jdiskmark/BenchmarkWorker.java Outdated Show resolved Hide resolved

jamesmarkchan reviewed Dec 7, 2025

View reviewed changes

src/jdiskmark/RunDetailsPanel.java

Copy link

Member

jamesmarkchan Dec 7, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nice this appears to be a pure swing panel, no generated code! awesome :)

jamesmarkchan reviewed Dec 7, 2025

View reviewed changes

src/jdiskmark/Util.java Outdated Show resolved Hide resolved

jamesmarkchan reviewed Dec 7, 2025

View reviewed changes

jamesmarkchan changed the title ~~[#95] Implement cache purging (admin + non-admin) with timing + logging~~ #95 Implement cache purging (admin + non-admin) with timing + logging Dec 7, 2025

vpstackhub added 4 commits December 8, 2025 11:56

Remove hard-coded project.dir from project.properties (#95)

47d2df6

Improve soft purge exception logging (Issue #95)

ce99e1e

Add logging for soft purge attempts (Issue #95)

ebd641d

#95 - Replace purge method strings with enum, fix UI panel, update CLI

e679860

jamesmarkchan reviewed Dec 15, 2025

View reviewed changes

jamesmarkchan force-pushed the dev branch from 028e039 to 34cc804 Compare December 17, 2025 06:25

Merge branch 'dev' into feature/cache-purge-95

a1d9ed1

jamesmarkchan reviewed Dec 28, 2025

View reviewed changes

jamesmarkchan assigned vpstackhub Dec 28, 2025

jamesmarkchan added 2 commits December 28, 2025 16:40

#95 minor house keeping

c02aca9

#95 allow project to build

a4c3e32

#95 Implement cache purging (admin + non-admin) with timing + logging #97

Are you sure you want to change the base?

#95 Implement cache purging (admin + non-admin) with timing + logging #97

Uh oh!

Conversation

vpstackhub commented Dec 5, 2025

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jamesmarkchan Dec 7, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jamesmarkchan commented Dec 7, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jamesmarkchan Dec 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

jamesmarkchan Dec 7, 2025 •

edited

Loading

jamesmarkchan Dec 28, 2025 •

edited

Loading