f



Re: raw devices in linux (O_DIRECT) #6



Martin Fuerderer escribi�:
> Hi,
> 
> in your Machine Specific Notes you should see something like this:
> 
> -------------------------------------------------------
> 3.  Kernel Asynchronous I/O (KAIO)
> 
>     Asynchronous I/O is supported by the official Linux kernel since 
> version
>     2.6.x. IBM Informix Dynamic Server supports Kernel Asynchronous I/O 
> (KAIO)
>     on character devices (a.k.a. raw devices) and block devices. It is 
                                                 ^^^^

> enabled
>     by default, and can be disabled by setting the environment variable
>     KAIOOFF=1 in the environment of the process that brings up the server.
> 
>     ...
> 
>     Note that the libaio.so library is required, regardless of the 
> distribution
>     or kernel level. At the time of writing, IBM Informix Dynamic Server
>     requires a minimum of libaio 0.3.96-3.
> -------------------------------------------------------
> 
> Where the last paragraph is exactly what Sandor has pasted into the
> e-mail before.
> 
> Which information do you think is missing ? Can you be more specific ?
> 

I's been using IDS on linux for years and always need to link the disk partitions (block 
devices) to raw devices (/dev/raw) to optimize I/O.

It wasn't clear to me, that I don't need any longer to use /dev/raw, and can just use the 
block device.

Theere's also no reference to O_DIRECT  in the release/en_us/0333/* files nor the 
Information Center site (http://publib.boulder.ibm.com/infocenter/ids9help/index.jsp)

----

Thanks everyone wich help me undertand how KAIO works now in kernel 2.6 and IDS 10.

IDS 10.0UC1 working fine in Fedora Core 3

Regards,


sending to informix-list
0
Fernando
3/4/2005 5:01:44 PM
comp.databases.informix 16083 articles. 0 followers. Post Follow

0 Replies
487 Views

Similar Articles

[PageSpeed] 24

Reply:

Similar Artilces:

Re: raw devices in linux (O_DIRECT)
Hi, On Thursday 03 March 2005 18:45, you wrote: > Hi, > > I just installed IDS 10 on linux and this is what I have found: > > - It required glibc 2.3.3 and I was using and old redhat 8, so I upgraded > linux to Fedora Core 3. Dependency fixed > > - Required libstdc++.so.5, FC3 includes libstdc++.6[.0.3] so make a > symbolic link (yes I should downgrade but like with curses4 works fine) Fedora Core 3 delivers compat-libstdc++-8-3.3.4.2.i386.rpm which include the right link to libstdc++.so.5. If you install the package you can resolve the dependency clean. Andreas > > - I can't use raw devices with FC3! it's depretacted, the new way is > using the O_DIRECT flag when opening the chunks, of course in a > close-source program like IDS the only hope is to open a feature request. > > + RHEL 4 has this deprecated options enabled so I don't think IBM > will be interested in this solution. > > + The advantage of O_DIRECT is you dont need to make the association > with the /dev/raw devices and rawio will be always in effect, also KAIO is > enabled by default with rawdevices. > > + The workaround is make a new linux-util and kernel package with raw > enabled. > > I'll make my custom kernel and try. > sending to informix-list sending to informix-list ...

Re: raw devices in linux (O_DIRECT) #5
Hi, in your Machine Specific Notes you should see something like this: ------------------------------------------------------- 3. Kernel Asynchronous I/O (KAIO) Asynchronous I/O is supported by the official Linux kernel since version 2.6.x. IBM Informix Dynamic Server supports Kernel Asynchronous I/O (KAIO) on character devices (a.k.a. raw devices) and block devices. It is enabled by default, and can be disabled by setting the environment variable KAIOOFF=1 in the environment of the process that brings up the server. ... Note that the libaio.so library is required, regardless of the distribution or kernel level. At the time of writing, IBM Informix Dynamic Server requires a minimum of libaio 0.3.96-3. ------------------------------------------------------- Where the last paragraph is exactly what Sandor has pasted into the e-mail before. Which information do you think is missing ? Can you be more specific ? Regards, Martin -- Martin Fuerderer IBM Informix Development Munich, Germany Information Management owner-informix-list@iiug.org wrote on 04.03.2005 16:22:59: > Hi Sandor: > > Without the library libaio I got > > 08:32:46 libaio.so.1: cannot open shared object file: No such file or > directory > 08:32:46 Please install libaio package. See machine notes. > 08:32:46 Warning: Kernel Asynchronous I/O is disabled. > > After installing the libr...

Re: raw devices in linux (O_DIRECT) #4
Hi Andreas, Andreas Breitfeld escribi�: > Hi, > > On Thursday 03 March 2005 18:45, you wrote: > >>Hi, >> >>I just installed IDS 10 on linux and this is what I have found: >> >>- Required libstdc++.so.5, FC3 includes libstdc++.6[.0.3] so make a >>symbolic link (yes I should downgrade but like with curses4 works fine) > > Fedora Core 3 delivers compat-libstdc++-8-3.3.4.2.i386.rpm which include the > right link to libstdc++.so.5. If you install the package you can resolve the > dependency clean. > > Andreas > With this package works perfectly Thanks sending to informix-list ...

Re: raw devices in linux (O_DIRECT) #3
Hi Sandor: Without the library libaio I got 08:32:46 libaio.so.1: cannot open shared object file: No such file or directory 08:32:46 Please install libaio package. See machine notes. 08:32:46 Warning: Kernel Asynchronous I/O is disabled. After installing the library and leaving raw devices disabled in kernel (default), IDS 10 uses KAIO. [root@tina 0333]# cat /proc/sys/fs/aio-nr 32768 [root@tina ~]# onstat -g iov IBM Informix Dynamic Server Version 10.00.UC1 -- On-Line -- Up 01:21:20 -- 1799820 Kbytes AIO I/O vps: class/vp s io/s totalops dskread dskwrite dskcopy wakeups io/wup errors kio 0 i 0.0 119 119 0 0 275 0.4 0 kio 1 i 0.0 195 185 10 0 502 0.4 0 kio 2 i 0.0 40 23 17 0 170 0.2 0 msc 0 i 0.0 13 0 0 0 14 0.9 0 aio 0 i 0.0 161 16 8 0 161 1.0 0 aio 1 i 0.0 2 0 0 0 3 0.7 0 aio 2 i 0.0 0 0 0 0 1 0.0 0 pio 0 i 0.0 1 0 0 0 2 0.5 0 pio 1 i 0.0 0 0 0 0 1 0.0 0 lio 0 i 0.0 0 0 0 0 1 0.0 0 There is no more information on release notes but machine notes mention the support of KAIO in kerne...

Re: raw devices in linux (O_DIRECT) #2
Hi Fernando, I have some information regarding the O_DIRECT flag. IDS V10 detects automatically that it runs on Linux Kernel 2.6 and if KAIO is enabled then it turns on the O_DIRECT flag! To enable KAIO: The libaio.so library is required, regardless of the distribution or kernel level. At the time of writing, IBM Informix Dynamic Server V10.0 requires a minimum of libaio 0.3.96-3. Please also read the release notes for more information! bye Sandor IBM Informix Development Munich Information Management Hollerithstra�e 1 81829 Munich Germany Fernando Ortiz <fortiz@lacorona. com.mx> To Sent by: "Informix-Discussion E-List owner-informix-li (E-mail)" <informix-list@iiug.org> st@iiug.org cc Subject 03.03.2005 18:45 raw devices in linux (O_DIRECT) ...

raw devices in linux (O_DIRECT)
Hi, I just installed IDS 10 on linux and this is what I have found: - It required glibc 2.3.3 and I was using and old redhat 8, so I upgraded linux to Fedora Core 3. Dependency fixed - Required libstdc++.so.5, FC3 includes libstdc++.6[.0.3] so make a symbolic link (yes I should downgrade but like with curses4 works fine) - I can't use raw devices with FC3! it's depretacted, the new way is using the O_DIRECT flag when opening the chunks, of course in a close-source program like IDS the only hope is to open a feature request. + RHEL 4 has this deprecated options enabled so I don't think IBM will be interested in this solution. + The advantage of O_DIRECT is you dont need to make the association with the /dev/raw devices and rawio will be always in effect, also KAIO is enabled by default with rawdevices. + The workaround is make a new linux-util and kernel package with raw enabled. I'll make my custom kernel and try. sending to informix-list ...

RE: Mirror / Raw device / Informix
This is a multi-part message in MIME format. ------_=_NextPart_001_01C6153D.0A4774A2 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Muthana- In order to use redundant drives on your server: in your onconfig file, change MIRROR 0 to MIRROR 1, then bounce your database. You may then use onspaces -m to mirror your dbspaces. Read the manual first. For replication across servers, Informix has built in high availability data replication (master - slave only) and enterprise replication (can be configured in many different ways). Both are free in enterprise edition. High availability requires an additional license for workgroup edition. Enterprise replication is not allowed in workgroup edition. For documentation, see: http://www.ibm.com/software/data/informix/pubs/library =20 --EEM =20 -----Original Message----- From: Muthana.AL-Temimi@tu-harburg.hamburg.de [mailto:Muthana.AL-Temimi@tu-harburg.hamburg.de]=20 Sent: Monday, January 09, 2006 10:06 AM To: informix-list@iiug.org Cc: m.al@tuhh.de Subject: Mirror / Raw device / Informix =20 Dear IIUG, =20 I would like to ask if I can do a mirror of my database which is installed on raw device into the 2 server by using mirror technology. I heart that this solution won't work with Informix but I'm not sure therefore I asked the questi...

RE: Relocate Informix database from Raw slices to filesystem
In the past we have used Informix's internal mirroring capabilities to move to new disk arrays. This should also work to move from raw to cooked disks. -----Original Message----- From: owner-informix-list@iiug.org [mailto:owner-informix-list@iiug.org]On Behalf Of mccmx@hotmail.com Sent: Monday, October 17, 2005 9:26 AM To: informix-list@iiug.org Subject: Relocate Informix database from Raw slices to filesystem Is there any way I can relocate my IDS 7.31 database from its raw slices to the local filesystem...? I can 'dd' the slices to files easily, but I need a way of telling Informix the new location of the chunks. This seems straigtforward for the rootdbs (which is defined in onconfig), but what about the other chunks..? I want to avoid unloading and reloading the data. All I want to do is move the physical chunks.. onstat -d: Informix Dynamic Server Version 7.31.UC5 -- On-Line -- Up 2 days 14:38:28 -- 40768 Kbytes Dbspaces address number flags fchunk nchunks flags owner name c04a150 1 1 1 1 N informix rootdbs c04a618 2 1 2 1 N informix logicdbs c04a6d8 3 1 3 1 N informix procdbs 3 active, 2047 maximum Chunks address chk/dbs offset size free bpages flags pathname c04a210 1 1 50 400000 278361 PO- /dev/rdsk/c1t6d0s4 c04a458 2 2 50 200000 134867 P...

Binding raw devices in Linux kernel 2.6
Hi, I'm trying to port an old program compiled on kernel 2.4, onto kernel 2.6. I've using a Mandriva 2009.1 box as my development PC. I've managed to get the program to build properly once I've gotten the necessary libraries, and it also runs. However, one of the first things that the program does is to bind the raw devices(already set up as separate logical partitions within the drive as /dev/sdaN) to /dev/rawN, in order to access them later directly. To do so, the first thing that it does is to open /dev/rawctl with the following code: [code] if ((master_fd = open("/...

RE: raw vs. cooked files under linux #6
When you compare cooked versus raw here, are you focusing on Linux only, or talking about this in general - other Unix platforms as well ? -----Original Message----- From: owner-informix-list@iiug.org [mailto:owner-informix-list@iiug.org] On Behalf Of Data Goob Sent: 11 May 2005 03:08 PM To: informix-list@iiug.org Subject: Re: raw vs. cooked files under linux If you are in a slow-change environment raw-files are not a bad thing. They create extra problems on the management side that cooked files do not. Fast-changing environments create more demand on your time so why put yourself into extra work with raw files? They don't give you enough of a difference in performance to really warrant them unless you are really demanding that extra 10%. Cooked files are also also more flexible on backups, you can back the dbspaces up with a regular backup tool or use the Informix backup tools, but raw files will have to be backed up with Informix tools only. SLES9 is quite a sea change from SLES8, and you could probably get away without SLES and go with the workstation Pro 9.3 if you are on a budget. SLES has better support for SCSI than the workstation versions and you must buy support or the online update doesn't work. It's an enterprise lock-in you might not need so try before you buy. SLES9 can be downloaded for trial, so give it a spin. You should be able to download IDS 10 for trial as well, thus no out of pocket to try it before pay...

Re: Informix and Linux opinions/who is running production #6
I know a few, but they are using Hydra from Querix. On 12 Feb 2004 at 15:16, Mario Canto wrote: > And... are there any linux-informix shop runing 4GL RDS applications on > linux? > Which informix versions, which linux platform? > It seems to me that the answer to this question is NULL. (I hope I am > wrong) > > Mario R. Canto > > Andrew Hamm wrote: > > > > Gary Quiring wrote: > > > I would like to know if we have any shops out here running IDS 9.x on > > > Linux in a production environment where you have at least...

RE: [SPAM] Re: Informix for linux
Because the IIUG server was out due to power from the hurricane that hit the East Coast of the United States The IIUG server is now back up and running At 09:16 PM 5/27/2003 -0400, Rob Vorbroker wrote: >ikke - I'm cc'ing this to the webmaster@iiug.org. >His name is Stuart Litel and once you know him...well. > >Perhaps he can figure out why you haven't recieved a >confirmation email in 3 days. Please remember that >everyone who is on the IIUG BoD or other committees >are volunteers. (: Kinda like IBM/Informix Marketing >I guess. LOL...

Re: Relocate Informix database from Raw slices to filesystem #4
True there are issues but it's a easy solution to the problem, gets round the 'can not rename chunks' issue, no need to hassle about mirroring etc But the original post said going to internal disks so there will not be a name clash. On 10/17/05, Paul Watson <pwatson@irace.com> wrote: > > On 17 Oct 2005 15:26:36 -0700, david@smooth1.co.uk <david@smooth1.co.uk> > wrote: > > > > > > Paul Watson wrote: > > > If the device is 'not going to exist anymore' why not dd the data from > > the > > > old slice to the internal disk and then > > > just link the 'not going to exist anymore device' to the new location > > > > > > > > > > > > I'm guess this a Solaris platform..and then someone runs devfsadm that > > rebuilds /dev and removes your links! > > > > Or someone adds a new device and Solaris wants to use that name! > > > > Bad idea to store stuff under /dev. That directory is for the OS > > to control not you. > > > > If you are running on IDS 9.40 or IDS 10 then ontape has the ability to > > rename chunks during a restore. > > > > http://www-128.ibm.com/developerworks/db2/library/techarticle/dm-0405fan/ > > > > > > > > > -- > Paul Watson > Tel: +44 7818 003457 > Fax: +44 1436 678693 > ...

Re: Relocate Informix database from Raw slices to filesystem #3
If the device is 'not going to exist anymore' why not dd the data from the old slice to the internal disk and then just link the 'not going to exist anymore device' to the new location On 10/17/05, Fernando Nunes <spam@domus.online.pt> wrote: > > Clive Eisen wrote: > > mccmx@hotmail.com wrote: > > > >> > >> I don't neccesarily need to move them off raw slices, but I do need to > >> move them to different slices. This is because I am moving the > >> database from SAN storage to internal disks... So ultimately > >> /dev/rdsk/c1d6d0s4 (and s5 and s6) wont exist anymore. > >> > >> So I need a way of relocating the data to a new slice or a filesystem > >> file. I suspect the process will be t > >> > > You can, in an unsupported environment, binary edit the rootdbs > > as the path to the other chunks is held there > > But the new paths would need to have the same length... > That's why you should never use physical names. Always use symbolic > links to the real devices... > > In this circunstances I don't see how an unload/reload can be avoided... > > > Regards. > > > -- Paul Watson Tel: +44 7818 003457 Fax: +44 1436 678693 sending to informix-list Paul Watson wrote: > If the device is 'not going to exist anymore' why not dd the data from the ...

Re: Relocate Informix database from Raw slices to filesystem #2
Why pay someone ??? It's a trivial task On 10/17/05, Neil Truby <neil.truby@ardenta.com> wrote: > > <mccmx@hotmail.com> wrote in message > news:1129561344.240798.314750@g49g2000cwa.googlegroups.com... > > Thats sounds promising. > > > > So effectively add a mirrored chunk to each dbspace using onspaces and > > then drop the original chunk. > > > > Can you do this for the rootdbs dbspace too...? > > > > Matt > > > > If it's a production system you might be better off paying someone who > knows > what they're doing to do this for you. > > -- > Neil Truby t:01932 724027 > Director m:07798 811708 > Ardenta Limited e:neil.truby@ardenta.com > > > > -- Paul Watson Tel: +44 7818 003457 Fax: +44 1436 678693 sending to informix-list ...

RE: Relocate Informix database from Raw slices to filesystem #5
-----Original Message----- From: owner-informix-list@iiug.org [mailto:owner-informix-list@iiug.org]On Behalf Of mccmx@hotmail.com Sent: Tuesday, October 18, 2005 3:59 AM To: informix-list@iiug.org Subject: Re: Relocate Informix database from Raw slices to filesystem Funnily enough, your idea to create a soft link in /dev from the old device to the new device/location is exactly what I was planning to do if I couldn't find a 'better' solution. Based on this thread, I may well have to go down that route... My Informix knowledge is very limited....!?! What I will probably do is: 1. create the soft link from /dev to get informix up and running on a different machine. 2. On this new machine: Add a mirror chunk to each dbspace and drop the original mirror to effectively relocate the databse to the filesystem. If I have no joy with step 2, I will run with the soft link from /dev. This will have to be documented well so that if someone destroys /dev (through devfsadm or similar) then the links can be recreated easily. Its not ideal but its preferable to unloading and reloading the whole database, which has its own risks. Thanks for the advice.. Matt onbar has a -rename option... sending to informix-list ...

RE: Re: Informix and SQL Server #6
we have a similar situation, we have not installed any unix odbc drivers on our unix machine, this is what i do, which works pretty well, - created import dts package to pull the informix tables in sqlserver with the following settings for Data Source "OLE DB Provider for ODBC Driver" on windows (not on unix) and used informix odbc driver which i believe comes with the sqlserver engine - run the dts package just before running the query the dts import packages pulls informix data and loads it to the sqlserver database at the rate of around 3000 rows/second (which is not too...

RE: database restore with ontape -r on Informix Online Dynamic Server 7.31 on Linux
David Williams wrote > > "Andy Kent" <andykent.bristol@virgin.net> wrote in message > news:a3525a11.0401080308.64a1ed12@posting.google.com... > > Why ever would you want to run a cold restore unattended? > > > > Great laxative ... > > > > To create a reporting database once per week... > > Once all the ontape stuff is finished don't forget to > > a) Sit in a loop running onstat - and check the return > code each time.Wait for the server to reach quiescent > mode. This can take 45 mins. >...

Re: hi Cant load informix online 5.2 on linux -8 ( rpm file) #6
hello friends. How are you. As i said. I did open and loaded the rpm file into /opt/infomix directory. and then copied the all contents to /usr/infomrix directory. From /usr/infomix i run from root . / installonline and the online started installion with some warnings and finally displayed the messaged that installation comeplete sucessfully. But when logingin into usr/infomix bu user infomrix. i cannt run any command like tbmonitor. dbaccess etc. it gives me the errore messeager like " SEGMENTATION FAULT" AND DO NOTHING., wHAT IS THIS. as the installion di...

RE: database restore with ontape -r on Informix Online Dynamic Server 7.31 on Linux #2
Christian Knappke wrote > > "Colin Bull" <c.bull@videonetworks.com> wrote: > > > Here is one I prepared earlier --- > > > > onbar/ontape whatever .. > > FORCEOL=n > > until onstat - | grep "On-Line" > /dev/null > > do > > echo "Off line" # ; sleep 20 > > onstat - | grep "[BQ][lu][oi][ce][ks]" > /dev/null && ( echo > > "\nServ blokd" > > if [ x$FORCEOL = "xn" ] ; then > > echo "On...

Re: database restore with ontape -r on Informix Online Dynamic Server 7.31 on Linux #3
Thanks a lot ! My problem was the fact that I tried to restore on a Linux machine a backup made on a UNIX machine. The INFORMIX server version was the same. The problem was the OS, in fact. Backup with ontape is recommended only when the restore is made on the same machine. Carmen ----- Original Message ----- From: "S.Dhayanidhi" <sdhaya@hotmail.com> To: <informix-list@iiug.org> Sent: Monday, January 12, 2004 6:35 AM Subject: Re: database restore with ontape -r on Informix Online Dynamic Server 7.31 on Linux > Here is my cute little code.. for u...

Re: Re: Informix and Linux opinions/who is running production
I know one. I've posted the name privately, as the last time I made the error of sharing information about a client on this list, a consultant (?) from this list cold called them. Nice behaviour. Pete _________________________________________________________________ Tired of 56k? Get a FREE BT Broadband connection http://www.msn.co.uk/specials/btbroadband sending to informix-list ...

Getting name of device which is bind to raw device if i have raw device name
Hi Can any one tell me Ihave a raw device is bind to a logical volume How to get the logical volume name from the name of the raw device Thanks , Sachin sachin wrote: > > Hi > Can any one tell me > Ihave a raw device is bind to a logical volume > How to get the logical volume name from the name of the raw device That is imposible. You can get the device number by using the RAW_GETBIND ioctl (like raw -q does). But there is no way to get the device name from the device number (other than searching through all of /dev and hoping to find the right nod...

How to copy a sun-database in one informix database to another informix database?
Hi, Sorry I am not an informix dba, and I was faced to copy a sub-database(sorry I don't know the clear name of such concept, as informix is like sql-server, which used multiple database) from one informix database system to another informix database. I know in oracle, we can use transportable tablespace, copy the datafile and transport the metadata, and do something else. How can I do it in informix then? Is there document which talk about the detail step? Thanks It would really help if you knew the versions of the Informix databases. Assuming Informix IDS 7.x or IDS...

Web resources about - Re: raw devices in linux (O_DIRECT) #6 - comp.databases.informix

Dan Rosenberg (@djrbliss) on Twitter
Sign in Sign up You are on Twitter Mobile because you are using an old version of Internet Explorer. Learn more here Dan Rosenberg @ djrbliss ...

ext4 - Wikipedia, the free encyclopedia
and proposed to fork the source code of ext3, rename it as ext4, and do all the development there, without affecting the current ext3 users. ...

on io scheduling again
Most of database engines have to deal with underlying layers - operating systems, device drivers, firmware and physical devices, albeit different ...

An ode to running a database on bare metal
... the things they recommended, which we implemented: set vm.swappiness kernel setting to 0 in /etc/sysctl.conf set InnoDB flush method to O_DIRECT ...

Subtly Bad Things Linux May Be Doing To PostgreSQL
In addition to talking about PostgreSQL at LSF/MM and Collab , I also learned a few things about the Linux kernel that I had not known before, ...

FastVPS Eesti OU · GitHub
Build software better, together.

Lucene and fadvise/madvise
While indexing, Lucene periodically merges multiple segments in the index into a single larger segment. This keeps the number of segments relatively ...


Disaster: MySQL 5.5 Flushing
We raised topic of problems with flushing in InnoDB several times, some links: InnoDB Flushing theory and solutions MySQL 5.5.8 in search of ...

Wim Coekaerts Blog
Blogs.Oracle.Com - Wim Coekaerts Blog

Resources last updated: 3/1/2016 6:18:40 PM