Introduce FORCE_DEFRAG compilation option to allow activedefrag run when allocator is not jemalloc #1303

ranshid · 2024-11-14T10:46:43Z

Introduce compile time option to force activedefrag to run even when
jemalloc is not used as the allocator.
This is in order to be able to run tests with defrag enabled
while using memory instrumentation tools.

fixes: #1241

Introduce compile time option to force activedefrag to run even when jemalloc is not used as the allocator. This is in order to be able to run tests with defrag enabled while using memory instrumantation tools. Signed-off-by: ranshid <[email protected]>

Signed-off-by: ranshid <[email protected]>

…efrag

codecov · 2024-11-14T12:00:16Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 70.87%. Comparing base (1acf7f7) to head (5ddb45e).

Additional details and impacted files

@@             Coverage Diff              @@
##           unstable    #1303      +/-   ##
============================================
+ Coverage     70.71%   70.87%   +0.15%     
============================================
  Files           119      119              
  Lines         64652    64616      -36     
============================================
+ Hits          45717    45794      +77     
+ Misses        18935    18822     -113

Files with missing lines	Coverage Δ
src/allocator_defrag.c	`100.00% <100.00%> (ø)`
src/config.c	`77.59% <ø> (-0.06%)`	⬇️
src/defrag.c	`88.80% <ø> (-0.23%)`	⬇️
src/server.h	`100.00% <ø> (ø)`

... and 45 files with indirect coverage changes

Signed-off-by: Ran Shidlansik <[email protected]>

madolson · 2024-11-15T05:07:33Z

I think we'll want to merge this PR first btw, #1242, since it already exists. I think there will be some merge conflicts.

CMakeLists.txt

madolson

Mostly makes sense to me. Some ideas for clarity but nothing to really change the direction.

src/CMakeLists.txt

madolson · 2024-11-15T05:32:02Z

src/defrag.c

@@ -755,6 +755,15 @@ void defragScanCallback(void *privdata, const dictEntry *de) {
 * or not, a false detection can cause the defragmenter to waste a lot of CPU
 * without the possibility of getting any results. */
 float getAllocatorFragmentation(size_t *out_frag_bytes) {


Kind of feels like you should override computeDefragCycles instead, and have that should always set active_defrag_running to like 100%.

I am fine with it (makes sense) I do not, however want to break tests which are checking the active_defrag_running, so maybe I can just use the active-defrag-cycle-max as the return value?
Alternatively we can have a tag to skip tests in case of DEBUG_FORCE_DEFRAG

Alternatively we can have a tag to skip tests in case of DEBUG_FORCE_DEFRAG

This intuitively makes a bit more sense to me, since defrag shouldn't really work correctly since we are completely breaking defrag here.

All of the tests already are checking for jemalloc memory to enable the defragmentation, so I think that the tests should all still be skipped anyways?

I think my last change keeps the same logic and the code is cleaner. I find having the defrag reporting the fragmentation based on the config limits is fine but take a look and let me know.

src/defrag.c

madolson · 2024-11-22T01:07:31Z

@ranshid Zvis change is now merged. Jim's is basically also ready to go so you should be able to resume updating this if you want.

src/defrag.c

ranshid · 2024-12-08T15:26:50Z

@ranshid Zvis change is now merged. Jim's is basically also ready to go so you should be able to resume updating this if you want.

Thank you. just got time to circle back on this one. merging and continue where I left off

Signed-off-by: Ran Shidlansik <[email protected]>

ranshid · 2024-12-09T09:43:20Z

we will probably hit issues from day 1. eg:

Logged sanitizer errors (pid 53103):
=================================================================
==53103==ERROR: AddressSanitizer: heap-use-after-free on address 0x6080000026b8 at pc 0x0001006130d0 bp 0x00016fa72ec0 sp 0x00016fa72eb8
READ of size 4 at 0x6080000026b8 thread T0
    #0 0x1006130cc in defragStageKvstoreHelper+0x650 (valkey-server:arm64+0x1002870cc)
    #1 0x100610d64 in activeDefragTimeProc+0x2e0 (valkey-server:arm64+0x100284d64)
    #2 0x1003c128c in aeProcessEvents+0x630 (valkey-server:arm64+0x10003528c)
    #3 0x10040334c in main+0x6d7c (valkey-server:arm64+0x10007734c)
    #4 0x19d1ff150 in start+0x9a8 (dyld:arm64e+0xfffffffffff4d150)

0x6080000026b8 is located 24 bytes inside of 96-byte region [0x6080000026a0,0x608000002700)
freed by thread T3 here:
    #0 0x101338d40 in free+0x98 (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x54d40)
    #1 0x1005bed10 in lazyfreeFreeDatabase+0x54 (valkey-server:arm64+0x100232d10)
    #2 0x100578750 in bioProcessBackgroundJobs+0x4e8 (valkey-server:arm64+0x1001ec750)
    #3 0x101335858 in asan_thread_start(void*)+0x40 (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x51858)
    #4 0x19d589f90 in _pthread_start+0x84 (libsystem_pthread.dylib:arm64e+0x6f90)
    #5 0x19d584d30 in thread_start+0x4 (libsystem_pthread.dylib:arm64e+0x1d30)

previously allocated by thread T0 here:
    #0 0x101338fd0 in calloc+0x9c (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x54fd0)
    #1 0x1004120a0 in valkey_calloc+0x3c (valkey-server:arm64+0x1000860a0)
    #2 0x1003cd844 in kvstoreCreate+0xa8 (valkey-server:arm64+0x100041844)
    #3 0x1005bf778 in emptyDbAsync+0xbc (valkey-server:arm64+0x100233778)
    #4 0x100458ce8 in emptyData+0x354 (valkey-server:arm64+0x1000ccce8)
    #5 0x100459cd4 in flushallCommand+0x1c0 (valkey-server:arm64+0x1000cdcd4)
    #6 0x1003e9500 in call+0x3cc (valkey-server:arm64+0x10005d500)
    #7 0x1003ec7a4 in processCommand+0x17b4 (valkey-server:arm64+0x1000607a4)
    #8 0x100424358 in processInputBuffer+0x498 (valkey-server:arm64+0x100098358)
    #9 0x100422204 in readQueryFromClient+0xd4 (valkey-server:arm64+0x100096204)
    #10 0x100668688 in connSocketEventHandler+0x160 (valkey-server:arm64+0x1002dc688)
    #11 0x1003c0fb8 in aeProcessEvents+0x35c (valkey-server:arm64+0x100034fb8)
    #12 0x10040334c in main+0x6d7c (valkey-server:arm64+0x10007734c)
    #13 0x19d1ff150 in start+0x9a8 (dyld:arm64e+0xfffffffffff4d150)

Thread T3 created by T0 here:
    #0 0x1013301c8 in pthread_create+0x5c (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x4c1c8)
    #1 0x100578104 in bioInit+0x28c (valkey-server:arm64+0x1001ec104)
    #2 0x1003ff6b4 in main+0x30e4 (valkey-server:arm64+0x1000736b4)
    #3 0x19d1ff150 in start+0x9a8 (dyld:arm64e+0xfffffffffff4d150)

I think it is after the changes introduced in #1242 since the db expires can be freed in the background but is passed as target in beginDefragCycle.

I wonder if to invest time to fix it now or address them later

Signed-off-by: Ran Shidlansik <[email protected]>

src/zmalloc.c

Signed-off-by: Madelyn Olson <[email protected]>

.github/workflows/daily.yml

madolson

Direction looks fine to me, but just a minor comment about removing one of the test permutations.

src/allocator_defrag.h

Signed-off-by: Madelyn Olson <[email protected]>

Co-authored-by: Madelyn Olson <[email protected]> Signed-off-by: ranshid <[email protected]>

…ests Signed-off-by: Ran Shidlansik <[email protected]>

Signed-off-by: Ran Shidlansik <[email protected]>

ranshid added 8 commits November 13, 2024 19:48

Add Cmake support

31a04dc

Signed-off-by: ranshid <[email protected]>

Make defrag always active when forced

a42e565

Signed-off-by: ranshid <[email protected]>

Simplify the FORCE_DEFRAG Cmake logic

1f26c5d

Signed-off-by: ranshid <[email protected]>

order and fix some uneeded changes

f36ee08

Signed-off-by: ranshid <[email protected]>

make changes to SERVER_FLAGS

27f4290

Signed-off-by: ranshid <[email protected]>

Merge branch 'unstable', remote-tracking branch 'origin' into force-d…

c1862eb

…efrag

ranshid requested a review from madolson November 14, 2024 11:45

fix spell and format issues

4643d97

Signed-off-by: Ran Shidlansik <[email protected]>

madolson reviewed Nov 15, 2024

View reviewed changes

CMakeLists.txt Outdated Show resolved Hide resolved

madolson reviewed Nov 15, 2024

View reviewed changes

madolson mentioned this pull request Nov 18, 2024

Fix LRU crash when getting too many random lua scripts #1310

Merged

zuiderkwast mentioned this pull request Nov 18, 2024

Remove valkey specific changes in jemalloc source code #1266

Merged

xbasel reviewed Dec 1, 2024

View reviewed changes

src/defrag.c Outdated Show resolved Hide resolved

Merge remote-tracking branch 'origin/unstable' into force-defrag

5d65221

ranshid added 7 commits December 9, 2024 10:23

merge and fix some pr comments

1744aab

Signed-off-by: Ran Shidlansik <[email protected]>

add daily workflow test

14e102e

Signed-off-by: Ran Shidlansik <[email protected]>

fix compilation on x86

bd2b04a

Signed-off-by: Ran Shidlansik <[email protected]>

format fixes

6606bb6

Signed-off-by: Ran Shidlansik <[email protected]>

more formatt fighting

ec6ad04

Signed-off-by: Ran Shidlansik <[email protected]>

one last try to solve the format issue

cf5841f

Signed-off-by: Ran Shidlansik <[email protected]>

format

329167f

Signed-off-by: Ran Shidlansik <[email protected]>

format

05d3b03

Signed-off-by: Ran Shidlansik <[email protected]>

madolson reviewed Dec 9, 2024

View reviewed changes

src/zmalloc.c Outdated Show resolved Hide resolved

Update src/zmalloc.c

0c9f535

Signed-off-by: Madelyn Olson <[email protected]>

madolson reviewed Dec 9, 2024

View reviewed changes

.github/workflows/daily.yml Outdated Show resolved Hide resolved

madolson approved these changes Dec 9, 2024

View reviewed changes

madolson reviewed Dec 9, 2024

View reviewed changes

src/allocator_defrag.h Outdated Show resolved Hide resolved

madolson and others added 6 commits December 9, 2024 12:33

Update src/allocator_defrag.h

6cf641a

Signed-off-by: Madelyn Olson <[email protected]>

Update .github/workflows/daily.yml

51e83ad

Co-authored-by: Madelyn Olson <[email protected]> Signed-off-by: ranshid <[email protected]>

add debug-defrag flag for tests + debug_defarg::skip tag option for t…

5a836a1

…ests Signed-off-by: Ran Shidlansik <[email protected]>

fix bad global tag

cb1b2ce

Signed-off-by: Ran Shidlansik <[email protected]>

preset debug_defrag global fleg

0c1b7b0

Signed-off-by: Ran Shidlansik <[email protected]>

Merge remote-tracking branch 'origin/unstable' into force-defrag

5ddb45e

hpatro mentioned this pull request Dec 12, 2024

defrag: eliminate persistent kvstore pointer and edge case fixes #1430

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Introduce FORCE_DEFRAG compilation option to allow activedefrag run when allocator is not jemalloc #1303

Introduce FORCE_DEFRAG compilation option to allow activedefrag run when allocator is not jemalloc #1303

ranshid commented Nov 14, 2024

codecov bot commented Nov 14, 2024 •

edited

Loading

madolson commented Nov 15, 2024

madolson left a comment

madolson Nov 15, 2024

ranshid Nov 15, 2024

madolson Nov 15, 2024

madolson Nov 15, 2024 •

edited

Loading

ranshid Dec 9, 2024

madolson commented Nov 22, 2024

ranshid commented Dec 8, 2024

ranshid commented Dec 9, 2024 •

edited

Loading

madolson left a comment

Introduce FORCE_DEFRAG compilation option to allow activedefrag run when allocator is not jemalloc #1303

Are you sure you want to change the base?

Introduce FORCE_DEFRAG compilation option to allow activedefrag run when allocator is not jemalloc #1303

Conversation

ranshid commented Nov 14, 2024

codecov bot commented Nov 14, 2024 • edited Loading

Codecov Report

madolson commented Nov 15, 2024

madolson left a comment

Choose a reason for hiding this comment

madolson Nov 15, 2024

Choose a reason for hiding this comment

ranshid Nov 15, 2024

Choose a reason for hiding this comment

madolson Nov 15, 2024

Choose a reason for hiding this comment

madolson Nov 15, 2024 • edited Loading

Choose a reason for hiding this comment

ranshid Dec 9, 2024

Choose a reason for hiding this comment

madolson commented Nov 22, 2024

ranshid commented Dec 8, 2024

ranshid commented Dec 9, 2024 • edited Loading

madolson left a comment

Choose a reason for hiding this comment

codecov bot commented Nov 14, 2024 •

edited

Loading

madolson Nov 15, 2024 •

edited

Loading

ranshid commented Dec 9, 2024 •

edited

Loading