Workflow of bibo queries
|
11-07-2020, 10:49 PM
(This post was last modified: 08-13-2021 06:31 PM by lingu.)
Post: #1
|
|||
|
|||
Workflow of bibo queries
The workflow of bibo queries use SageRun in Sage. SageRun drives an ATPA to perform a task for a QID.
The call path starts in scali invoking ProcRequestDefault. We optimize the timing of the steps and should make it complete in no more than 1s. r1: scali -> r3: getRequest -> r5: ProcRequestDefalt -> r7: queryMatch -> r9: SageRun instances [1..llTaskCnt] -> r11: atpa by user sage (atpa may log some info in atpa.log-$dd where dd is the datum number. Note that some log may be somewhere else.) -> r13: atpa.sh by user sage (log in atpash.log-$dd, also check output files in the query's dir in fastdata)) -> r15: procone.py (log is perhaps in atpa.sh's log or procone2-$tycano.log) Most of the log file locations are /thinker/local/today/users/$USER We optimized r1 to r7 from 20+s to 3s by reducing memory footprint used by SageRun, skipping SageRunPrep and increasing atp count. r9 and r11 still takes 13s. If we reduce the task count to 90, the time is reduced to 8s. With atpa.sh, it takes 10s. --- 20210813/lingu: log info 20210811/zhihao: link procone2.py 20201108/lingu: rename to be workflow. |
|||
11-08-2020, 01:39 AM
(This post was last modified: 11-08-2020 02:03 AM by lingu.)
Post: #2
|
|||
|
|||
RE: SageRun
A query may run in the following timing starting at 0025 for query 56.
25.45 issue: talk2thinker2 working for tok 002545292938784 [1108-00:25:45] 25.45 receive: received new case in /thinker/globe/soft/bibo/procuratorate/tests/newcase/newcase.txt [1108-00:25:45] 25.45 scali 25:47 thinkres start [2020-11-08 00:25:47] thinkres starts with --reqtype query --resdir /thinker/globe/soft/bibo/procuratorate/tests/res2 --casecnt 26 --out /thinker/globe/soft/bibo/procuratorate/tests/res2/res.res26.09 first wrid dir in qfastdir on limbo30-1 -- 24 26.09 first task 24 creates wrid dir in qfastdir 26.15 bibo.log last modification 26.29 last seen dir in qfastdir on limbo30-1 -- 126.seen |
|||
11-08-2020, 02:00 AM
Post: #3
|
|||
|
|||
RE: SageRun
I commented out the invokation of atpa.sh in atpa, and run the query again around 0054 with query 64.
54:29 issue 54:31 thinkres 54:51 modification of earliest dir 24 54.59 bibo.log final modification 55.07 modification on 126.seen. |
|||
11-08-2020, 02:34 AM
Post: #4
|
|||
|
|||
RE: SageRun
I try to reduce the amount of UPFS to be shared to others.
Run a query 0 around 0130 for query 16 without atpa.sh. 30:19 issue 30.31 first dir in qfastdir -- 24 30:44 last modified dir -- 126.seen 30:47 send results |
|||
11-08-2020, 02:44 AM
Post: #5
|
|||
|
|||
RE: SageRun
I add back the call to atpa.sh, and run a query 40 around 0140.
40:32 issue 40:34 thinres 40:45 first dir in qfastdir 41:05 last modified dir 126.seen |
|||
11-08-2020, 02:56 AM
(This post was last modified: 11-08-2020 05:00 AM by lingu.)
Post: #6
|
|||
|
|||
RE: SageRun
I reduce llUpfsSmall to 0x10000000 (64M) and run inquire_zeng around 0152 for query 0 -- but sage crashed.
I made it 256MiB and it seems to work. 52:21 issue 52:48 first dir in qfastdir 53:05 last modified dir 120.seen The perf is still bad. |
|||
11-08-2020, 05:05 AM
Post: #7
|
|||
|
|||
RE: SageRun
I further increased sage_atp_count to 30 and re-installed sage.
Code: [sage@limbo30-1 test]$ ./test_insert_zeng.sh Then I tested inquiry for qid 32 around 0402. Code: [sage@limbo30-1 test]$ ./test_inquire_zeng.sh 02:09 issue 02:12 thinkres starts 02:13.28 firest dir (3) in qfastdir 02.28:18 last modified dir (126.seen) in qfastdir 02:38 send results |
|||
11-08-2020, 10:03 AM
Post: #8
|
|||
|
|||
RE: SageRun
Tested query 40 without calling atpd.sh around 0853.
Code: sending inquire request ./processed/zengxingliang-inquire.json 53:57 issue 54:00 first dir 54:13 last dir |
|||
11-08-2020, 10:42 AM
(This post was last modified: 11-08-2020 10:48 AM by lingu.)
Post: #9
|
|||
|
|||
RE: Workflow of bibo queries
20201108 0935/lingu runs query 48 with sacli requesting 100 tasks without calling atpa.sh.
Code: [sage@limbo30-1 test]$ ./test_inquire_zeng.sh 35:43 issue 35:46 first fast dir 35:56 last fast dir Then I reduced task count to 90 around 0944. Code: [lingu@limbo30-1 res2]$ ll -tr 56|head -n2 44:46 issue 44:49 first fast dir 44:57 last fast dir |
|||
11-08-2020, 11:07 AM
Post: #10
|
|||
|
|||
RE: Workflow of bibo queries
20201108 1000/lingu runs query 64.
Code: [sage@limbo30-1 test]$ ./test_inquire_zeng.sh bibo.log shows the following. Code: received new case in /thinker/globe/soft/bibo/procuratorate/tests/newcase/newcase.txt [1108-10:00 00:28 issue 00:30 thinkres 00:32 first fast dir 00:42 last fast dir 00:57 send result |
|||
« Next Oldest | Next Newest »
|
- View a Printable Version
- Send this Thread to a Friend
- Subscribe to this thread
- Show the subscribers of this thread:
- Add subscribers to this thread: