Flexible File Filtering Feature for Blobfuse Mounting #1435

t-yashmahale · 2024-06-13T04:54:14Z

This pull request introduces advanced filtering capabilities to Blobfuse, empowering users to selectively mount cloud files based on size, regex patterns, modification time, and formats like PDF, JPG, TXT, and more. Efficient thread pool management at both file and filter levels ensures swift processing, enhancing user flexibility and optimizing resource utilization. This enhancement streamlines the file mounting process, enabling users to tailor their cloud storage access with precision and performance.

…ison

vibhansa-msft · 2024-06-13T04:56:47Z

.gitignore

@@ -21,3 +21,4 @@ lint.log
 azure-storage-fuse
 bfusemon
 test/scripts/dirIterate.go
+filterbin


There is this binary file _debug* which can be removed from the PR.

vibhansa-msft · 2024-06-13T04:58:11Z

component/azstorage/config.go

-	CPKEnabled              bool   `config:"cpk-enabled" yaml:"cpk-enabled"`
-	CPKEncryptionKey        string `config:"cpk-encryption-key" yaml:"cpk-encryption-key"`
-	CPKEncryptionKeySha256  string `config:"cpk-encryption-key-sha256" yaml:"cpk-encryption-key-sha256"`
+	BlobFilter              string `config:"blobFilter" yaml:"blobFilter"`


Move this new config towards the end of the structure so that diff does not show existing items.

vibhansa-msft · 2024-06-13T04:58:47Z

component/azstorage/config.go

-
+	az.stConfig.blobFilter = opt.BlobFilter
+	if len(opt.BlobFilter) > 0 {
+		fmt.Println(opt.BlobFilter)


You can move fmt.Println to log.Debug as blobfuse will have no console attached to print,

vibhansa-msft · 2024-06-13T04:59:29Z

component/azstorage/config.go

+		fmt.Println(len(filter.GlbFilterArr))
+		fmt.Println(len(opt.BlobFilter))
+		if erro != nil {
+			log.Err("no filters applied, mount failed")


For log.Debug/Err we are following a format which is consistent across all our logs. Just check other existing logs in this file and modify accordingly.

vibhansa-msft · 2024-06-13T04:59:40Z

component/azstorage/config.go

+	if len(opt.BlobFilter) > 0 {
+		fmt.Println(opt.BlobFilter)
+		erro := filter.ParseInp(&opt.BlobFilter)
+		fmt.Println(len(filter.GlbFilterArr))


These shall be converted to log.Debug or log.Info

vibhansa-msft · 2024-06-13T05:01:03Z

component/azstorage/azstorage.go

@@ -331,6 +333,10 @@ func (az *AzStorage) StreamDir(options internal.StreamDirOptions) ([]*internal.O
 	// increment streamdir call count
 	azStatsCollector.UpdateStats(stats_manager.Increment, streamDir, (int64)(1))

+	//check for filters provided
+	if len(az.stConfig.blobFilter) > 0 { //only apply filter if user has given


After parsing we might have created a filter object so instead of checking the string length here rather we can check if that object is nil or not, will be a faster check.

vibhansa-msft · 2024-06-13T05:01:20Z

component/azstorage/config.go

+			return erro
+		}
+	}
+	//blobfilter


Deadcode can be removed.

vibhansa-msft · 2024-06-13T05:02:05Z

component/azstorage/azstorage.go

+		return resp, err
+	}
+
+	if len(az.stConfig.blobFilter) > 0 {


Instead of checking length try to check object is nil or not.

vibhansa-msft · 2024-06-13T05:02:43Z

component/azstorage/azstorage.go

+
+	if len(az.stConfig.blobFilter) > 0 {
+		fv1 := &filter.FileValidator{
+			FilterArr: filter.GlbFilterArr,


Here we can name the variables better than fv1 and GlbFilterArr.

vibhansa-msft · 2024-06-13T05:03:41Z

component/azstorage/azstorage.go

@@ -331,6 +333,10 @@ func (az *AzStorage) StreamDir(options internal.StreamDirOptions) ([]*internal.O
 	// increment streamdir call count
 	azStatsCollector.UpdateStats(stats_manager.Increment, streamDir, (int64)(1))

+	//check for filters provided
+	if len(az.stConfig.blobFilter) > 0 { //only apply filter if user has given
+		new_list = filter.ApplyFilterOnBlobs(new_list)


Shall we keep the input and output variables same ? just think it might not be a safer way to do conversion.

vibhansa-msft · 2024-06-13T05:06:17Z

internal/filter/filter.go

+	"github.com/Azure/azure-storage-fuse/v2/internal"
+)
+
+var GlbFilterArr [][]Filter //it will store the fliters, outer array splitted by ||, inner array splitted by &&


As part of design we shall move these filters to be a class so that there can be multiple instances of such filters across the blobfuse code. Current design of having it as a global variable restricts that only one such filter can exists in scope of blobfuse instance.

vibhansa-msft · 2024-06-13T05:07:40Z

internal/filter/filter.go

+
+var GlbFilterArr [][]Filter //it will store the fliters, outer array splitted by ||, inner array splitted by &&
+
+func ApplyFilterOnBlobs(fileInfos []*internal.ObjAttr) []*internal.ObjAttr { //function called from azstorage.go streamDir func


From outside it shall be an object based call.

read config as a string from cli

parse it and create a filter object

filtering code shall be executed on this object

if filters are not applied then object shall be nil

if not nil then apply the filters and it will return back a new slice.

this model will allow us to have multiple instances of filter across the code if required,

souravgupta-msft · 2024-06-13T07:12:29Z

tools/filter/formatfilter.go

@@ -0,0 +1,34 @@
+package main


tools/filter can be removed

…ed to filter module

vibhansa-msft · 2024-06-14T10:48:41Z

component/azstorage/azstorage.go

+	//check for filters provided
+	if az.stConfig.filters != nil { //only apply if user has given filter
+		filtered_list := az.stConfig.filters.ApplyFilterOnBlobs(new_list)
+		// return filtered_list, *new_marker, nil


Better to return here instead of copying over the slice again.

…tomic variable

vibhansa-msft · 2024-06-19T10:36:20Z

internal/filter/accesstier.go

+func (filter AccessTierFilter) Apply(fileInfo *internal.ObjAttr) bool {
+	// fmt.Println("AccessTier filter ", filter, " file name ", (*fileInfo).Name)  DEBUG PRINT
+	fmt.Println("inside filter tier ", filter, " with given tier ", filter.tier, " and file tier ", fileInfo.Tier)
+	if (filter.opr == "=") && (filter.tier == strings.ToLower(fileInfo.Tier)) {


Multiple string checks can be avoided.

vibhansa-msft · 2024-06-19T10:36:37Z

internal/filter/accesstier.go

+
+func giveAccessTierFilterObj(singleFilter *string) (Filter, error) {
+	(*singleFilter) = strings.Map(StringConv, (*singleFilter)) //remove all spaces and make all upperCase to lowerCase
+	sinChk := (*singleFilter)[4:5]                             //single char after tier (ex- tier=hot , here sinChk will be "=")


Instead of magic numbers use len("keyword")

t-yashmahale added 24 commits May 29, 2024 10:49

Implemented 1st step that is parsing of filter string given by user

b88dc62

added a dummy threadpool where 16 threads are accessing files parallely

8d66bdb

added <= and >= features in size filter

b88e380

Added all functions for checking the file against filters

a3bd1d8

solved bug of deadlock if a file does not returns true for any compar…

05e6869

…ison

created modules for code and added meaningfull names

5107413

added comments for complex functions

f4b9ccf

added regex filter

99d3458

added modification time filter (currently supporting ust only)

9f54faa

made some changes to make code more efficient

c817cba

increased modularity

448e473

more optimised prev commit

02c70d5

included reading from output channel

7af33a5

passed all variables by refrence and other optimisations

1f290e4

started integration ,added input flag blobFilter

df32e9e

updated filter.go

024a9e3

integrated filters with blobfuse

aec9db8

filter integrated with prints

e19c97a

minor changes

27eca4f

_

057f84e

invalid input filter handeling done

c48ef28

GetAttr done and error handling in case of wrong filter

861a4a0

added comments

0555121

made some more changes to clean code

733bad0

t-yashmahale requested review from vibhansa-msft, souravgupta-msft and ashruti-msft as code owners June 13, 2024 04:54

vibhansa-msft reviewed Jun 13, 2024

View reviewed changes

component/azstorage/config.go Outdated

return erro

}

}

//blobfilter

Copy link

Member

vibhansa-msft Jun 13, 2024

Choose a reason for hiding this comment

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

Deadcode can be removed.

vibhansa-msft reviewed Jun 13, 2024

View reviewed changes

souravgupta-msft reviewed Jun 13, 2024

View reviewed changes

tools/filter/formatfilter.go Outdated

@@ -0,0 +1,34 @@

package main

Copy link

Member

souravgupta-msft Jun 13, 2024

Choose a reason for hiding this comment

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

tools/filter can be removed

vibhansa-msft linked an issue Jun 13, 2024 that may be closed by this pull request

Is it possible to mount only files which more than specific size.. In my case I want mount the files which is greater than 52B #1295

Open

t-yashmahale added 2 commits June 14, 2024 00:18

made changes

40dc026

solved bug of infinite waiting when a list containing no file is pass…

8d5cdb0

…ed to filter module

t-yashmahale requested review from vibhansa-msft and souravgupta-msft June 14, 2024 02:42

vibhansa-msft reviewed Jun 14, 2024

View reviewed changes

t-yashmahale added 3 commits June 14, 2024 20:11

resolved issue of infinite waiting in recieveOutput function due to a…

2532fa0

…tomic variable

added access tier filter

08c01f1

added directory functionality

cd8e469

vibhansa-msft reviewed Jun 19, 2024

View reviewed changes

t-yashmahale added 3 commits June 20, 2024 00:49

added blobtag filter and some other changes

7f98d4a

Enable read-only mode when user specifies filters

61450e3

changed name of flag to blob-filter

a5228d4

vibhansa-msft added V2 feature request labels Jul 23, 2024

vibhansa-msft added this to the v2-2.4.0 milestone Jul 30, 2024

vibhansa-msft removed this from the v2-2.4.0 milestone Nov 4, 2024

jainakanksha-msft added this to the v2-2.4.1 milestone Nov 4, 2024

vibhansa-msft merged commit 8fb1141 into Azure:feature/filter Dec 9, 2024
1 check was pending

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Flexible File Filtering Feature for Blobfuse Mounting #1435

Flexible File Filtering Feature for Blobfuse Mounting #1435

t-yashmahale commented Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

vibhansa-msft Jun 13, 2024

souravgupta-msft Jun 13, 2024

vibhansa-msft Jun 14, 2024

vibhansa-msft Jun 19, 2024

vibhansa-msft Jun 19, 2024 •

edited

Loading


		var GlbFilterArr [][]Filter //it will store the fliters, outer array splitted by \|\|, inner array splitted by &&

		func ApplyFilterOnBlobs(fileInfos []internal.ObjAttr) []internal.ObjAttr { //function called from azstorage.go streamDir func

Flexible File Filtering Feature for Blobfuse Mounting #1435

Flexible File Filtering Feature for Blobfuse Mounting #1435

Conversation

t-yashmahale commented Jun 13, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vibhansa-msft Jun 19, 2024 • edited Loading

Choose a reason for hiding this comment

vibhansa-msft Jun 19, 2024 •

edited

Loading