Re: qio, async_io and control files

andreyNSPAM_at_bookexchange.net (NetComrade) wrote in message news:<3f12eb82.600408281_at_nyc.news.speakeasy.net>...
>
> We use the qio product, and all data files are qio.
> trussing ora_<SID>_dbwr shows that most calls success as successful
> kaio calls (asyc_io is set to true)
> However, trussing ora_<SID>_ckpt shows that kaio calls fail 3 times in
> a row to different files, since there are 3 such calls and I have 3
> control files and control files are not qio files, I assume it's for
> control files (if anybody can teach me how to read parameters of kaio
> calls, i'd be very happy!)

...
>
>
> kaio(AIOWRITE, 299, 0x01A69EA0, 4096, 12288, 0x01A67EA8) Err#48
> ENOTSUP
> lwp_cond_signal(0xFEB21FA0) = 0
> lwp_cond_wait(0xFEB21FA0, 0xFEB21FB0, 0x00000000) = 0
> kaio(AIOWRITE, 298, 0x01A69EA0, 4096, 12288, 0x01A67DCC) Err#48
> ENOTSUP
> pwrite64(299, "1502\0\0\0\0\003\001 ]84".., 4096, 12288) = 4096
> lwp_cond_signal(0xFEB0FFA0) = 0
> lwp_cond_wait(0xFEB0FFA0, 0xFEB0FFB0, 0x00000000) = 0
> kaio(AIOWRITE, 297, 0x01A69EA0, 4096, 12288, 0x01A680F4) Err#48
> ENOTSUP
> pwrite64(298, "1502\0\0\0\0\003\001 ]84".., 4096, 12288) = 4096
>
> .......
> We use Oracle 8.1.7.4 on Solaris 2.7 boxes

Hi, NetComrade,

I can't answer other questions. But you don't have to guess about the file. lsof (freeware from sunfreeware.com) can tell you what file a process is processing on. Your truss output tells you the file descriptors 299 and 298. Then lsof -p <pid> -a -d 298,299 shows the file path.

The kaio system call is not documented. The arguments are operation type, file descriptor, pointer to the buffer where read data will be placed or data to be written is placed, read or write buffer size, offset where read or write begins. The last argument has something to do with read or write result. Not quite sure.

Yong Huang Received on Tue Jul 15 2003 - 14:37:59 CDT