[LU-14361] Add support for statahead pattern with regularized file name

• Status: Open
• Project: Lustre
• Type: Improvement
• Priority: Minor
• Reporter: Qian Yingjin
• Assignee: Qian Yingjin
• Resolution: Unresolved
• Labels: statahead

Description

Some applications do stat() calls under a directory within which all the children files have regularized file name:

• mdtest benchmark: mdtest.$i
• du more than 10000 entries: file.$i
• ML/AI with ingested data that have typically a rule of filename in the directory.

These applications call stat() on the files in the Alphabetic sorting order of the file name.

This kind of metadata operations can not be optimized by current statahead mechanism.
The current statahead mechanism works as follows:
1. Open the directory via opendir() call. It will authorize the statahead.
2. readdir() to get the name and inode number for the dentries.
3. do stat() on the dentries one by one.
4. Close the directory which will deauthorize the statahead.

In current statahead mechanism, the stat() calls in the order of populate the dentries via readdir(). For the ldiskfs backend, it is ordering by the hash of the file name, not a kinds fo sorting oder.

However, we can improve the statahead to support statahead pattern with regularized file name to optimize the metadata performance for the above applications

Issue Links

• is related to: LU-17399 sanity: test_123i Error: 'File name p...
• is related to: LU-14139 batched statahead processing
• is related to: LU-17450 sanity: interop test failures with ma...
• is related to: LU-14139 batched statahead processing

Comments

[Gerrit Updater]

Yingjin Qian ([e-mail]) uploaded a new patch: https://review.whamcloud.com/41308
Subject: LU-14361 statahead: regularized fname statahead pattern
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: f0027f94edf2d022dfcfe027ff1121e54f3ef345

[Gerrit Updater]

Yingjin Qian ([e-mail]) uploaded a new patch: https://review.whamcloud.com/41319
Subject: LU-14361 statahead: fname statahead pattern w/o opendir
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 8e5e9702b657cd354b00f58d5ad20b5324bb364f

[Gerrit Updater]

"Qian Yingjin <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/48625
Subject: LU-14361 statahead: add statahead support for mdtest
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 03a5ad01ecbf356b5d28580c36deb96f39ef6a19

[Gerrit Updater]

"Qian Yingjin <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/48954
Subject: LU-14361 statahead: add support for mdtest shared dir workload
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 377e3300cef3e297aab84bc317b3d6ad036f7167

[Andreas Dilger]

One of the questions I had recently from a customer is whether this statahead would work with a filename that had numbers in the middle (e.g. "filename.12345.out" or similar) because the filename extension is needed for some programs to work correctly. The current version of the patch does not handle this correctly, but I could imagine that this can is detected fairly easily. Instead of just comparing numbers from the end of the filename, it should be possible to build a "format" of characters that are changing between two filenames (the byte offset of the first numeric character and the length, and the file suffix or an empty string if no suffix).

The first filename would just be saved, then the second one would do a byte-by-byte comparison to find the first character that is different, then do a forward and backward search with isdigit() to find the first and last number in the filename. It would then use "snprintf(buf + start, sizeof(buf) "%*0u%s", len, index, suffix)" (or similar, with/without leading zeroes and fixed or variable width) to write the new index number into the middle (or end) of the filename.

[Gerrit Updater]

"Qian Yingjin <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/49291
Subject: LU-14361 statahead: wait inuse entry finished during cleanup
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 05d7d9fdc417fb9ddfec7379a70d5f318a2d1333

[Gerrit Updater]

"Qian Yingjin <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/51572
Subject: LU-14361 statahead: add tunable for fname pattern statahead
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 77f13ed7daa5858bc1a65a7ad782e6bdc6689981

[Gerrit Updater]

"Qian Yingjin <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/51587
Subject: LU-14361 statahead: return ENOENT for batched statahead
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 514378215024010c1a7318e01e201028ce56b7c4

[Gerrit Updater]

"Qian Yingjin <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/51592
Subject: LU-14361 statahead: add tunable for fname pattern detection
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 533efbf4c0f3ba13f7b18d6998c1d6f7f2e8738e

[Gerrit Updater]

"Qian Yingjin <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/51730
Subject: LU-14361 statahead: Add test for statahead advise
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: c774d17f9a572d01629f7cead0a126d88a9c290b

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/48625/
Subject: LU-14361 statahead: add statahead advise IOCTL
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 1288681bb8aa76f43af5c53ec3ef68ecc8d412d7

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/51730/
Subject: LU-14361 statahead: Add test for statahead advise
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 5317f8a523ac32b616713396f60eda4995dea85e

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/51572/
Subject: LU-14361 statahead: add tunable for fname pattern statahead
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 6fbffd9c09ad4526ce3a03b8d85f0ce273cac037

[Gerrit Updater]

"Timothy Day <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/53173
Subject: LU-14361 tests: add aheadmany to .gitignore
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 5027f6bc7fceee920ac5a54480db572dbfbcd5bd

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/53173/
Subject: LU-14361 tests: add aheadmany to .gitignore
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 46d69b57d4a2d9eeb4ac7adc472a7667394de919

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/41308/
Subject: LU-14361 statahead: regularized fname statahead pattern
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: e10bf68d7c3eb71a7e84986089c020aa4c9f98f7

[Andreas Dilger]

It looks like the newly-added sanity test_123i from patch https://review.whamcloud.com/41308 "LU-14361 statahead: regularized fname statahead pattern" is failing fairly often since the patch was landed (33/561 = 1/17 runs in the past week). Since the test is in sanity, it is run many times for each patch and much more likely to cause problems:

https://testing.whamcloud.com/search?horizon=2332800&status%5B%5D=FAIL&test_set_script_id=f9516376-32bc-11e0-aaee-52540025f9ae&sub_test_script_id=7d5be5fb-0bc8-4e25-a53f-3a833bca78bd&source=sub_tests#redirect

[Qian Yingjin]

The reason for these failures may be that @sbi->ll_sa_min is too small, the default value is 

#define LL_SA_REQ_MIN_DEF 8,

it results in the stat-ahead is not so aggressive, too low hit rate and quit wrongly...

Increasing this value in sanity/123i should fix the failure. 

 

I will make a patch later.

[Gerrit Updater]

"Qian Yingjin <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/53634
Subject: LU-14361 statahead: increase the initial statahead count
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: dd6b7c0f476085c0cbe0c0a4c05710e1205b5d25

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/53634/
Subject: LU-14361 statahead: increase the initial statahead count
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 76ff18798974b15c4a171dbb76a1077623a66d8a