Support query re hugepages in EC2 - Oracle on Amazon Web Services Cloud (Archived)

Hi, the XEN kernel in the OEL R5 AMIs does not seem to want to work with hugepages for large SGA database sizes; can we replace the kernel with a stock one ?

Related

HugePages on Suse Linux

On a Suse Linux 64 BIT System with 32GB RAM and 8 CPUs, I'll be installing 4 11gR2 Instances.
Will configuring HugePages on the O/S help in this situation?
If so, how do I determine what values should be allocated for all the HugePage parameters?
Thanks in advance. 
Help what in what situation?
http://docs.oracle.com/cd/B28359_01/server.111/b32009/appi_vlm.htm
Aren't you installing a 64bit version of Linux? 
I have a Suse Linux 64 BIT System with 32GB RAM and 8 CPUs. It has 4 11gR2 Instances installed.
Would configuing HugePages help performance with these 4 Instances? If so, how do I determine what values I should allow to HugePages? 
When people provide you with links there is generally value in reading them. 
I did read it and found nothing that applied that I could use. If there is, please specifiy. 
rewh2oman wrote:
I did read it and found nothing that applied that I could use. If there is, please specifiy.http://docs.oracle.com/cd/B28359_01/server.111/b32009/appi_vlm.htm#CACDCGAH 
You can either use kernel hugepages or shared memory, not both. A new feature of Oracle 11g is AMM (automatic memory management), which relies on /dev/shm shared memory under Linux. AMM is not compatible with kernel hugepages. Kernel hugepages is somewhat more efficient than AMM, which is basically software, but requires that you know what you are doing. AMM is default in 11g and certainly the the more convenient way. The performance gain of hugepages over AMM is probably not worth the additional hassle, but some Linux hardliner may see this differently.

any known issues using fstrim in oel 6.3 or 6.5 os ?

was there any known issues using fstrim in oel 6.3 or 6.5 os ? We would like to implement it either in fstab (with discard mount option ) or using cron job.
I guess this question is really about discard/trim support for SSD devices. According to my findings, to schedule a periodic "fstrim" action is the preferred or considered the safest solution.  Perhaps Avi Miller-Oracle or Tommyreynolds-Oracle from the Oracle Linux team can shred some more light on the subject, but I suggest to follow advice 18.6 SSD Configuration Recommendations for btrfs, ext4, and swap in the Oracle Linux documentation.  If you wish to schedule TRIM actions, I suggest the following, for example: 
/etc/cron.weekly/fstrim
#! /bin/sh
for mount in / /boot /home; do fstrim $mount; done
Of course you will have to substitute your SSD mount points (devcies) in the above example.Btw, OEL was renamed to "Oracle Linux" or OL with the introduction of the Oracle UEK kernel starting with OEL 5.5 and OL 5.6. I could not see any compelling reason to stick to 6.3 or 6.5 and not update these system to the 6.6, or whatever is currently the latest update level. 
Thanks for the info, but my question is was there any known issues by running fstrim on SSD mounts in oracle linux 6.5 or 5.5 os based hosts.
There are no known issues running fstrim on Oracle Linux 6. Oracle Linux 5 Update 5 is very, very old though and I don't even know if it properly supports SSDs or fstrim, given how old the kernel and hardware support is. You'd want to get up to Oracle Linux 5 Update 11 and run the UEK Release 2 at least before I'd be comfortable recommending SSDs and fstrim.
Awesome.Thanks for the quick update Avi. Lastly this fstrim is required for all ssd disks right ? I know it's not mandatory. but this will improve the disks performance.
You'd have to check with your SSD manufacturer and the device model. It's (sometimes) device dependent. Most SSDs should benefit from TRIM.

Redhat 5 has updated the kernel will there be an update to the oracleasm libraries

I know that the oracleasm libraries are kept in sync with updates from redhat 5 (el5, RHEL5) there is a new(410) version of the kernelkernel-2.6.18-410.el5.x86_64.rpmI can only find oracleasm-2.6.18-409.el5-2.0.5-1.el5.x86_64.rpmas the latest oracle asm library (From what I can tell updated in March)Will there be a new version coming out? I have not found any information searching the k-bases and community.Will the oracleasm-2.6.18-409.el5-2.0.5-1.el5.x86_64.rpmwork with the latest kernel update (410) from Redhat? thanksgary
Gary, This is what I got when I searched for the same. It seems that updates are released  for the most latest version of RH.http://www.oracle.com/technetwork/server-storage/linux/asmlib/rhel7-2773795.html Edit: So if you have the old binaries available from OTN, it should be working for RH5 . I have EL6 with me with the oracleasm package installed of the version 2.1.8-1.el6, don't have RH to check. HTHAman....
And this is why one  should be careful of tainting the kernel with vendor supplied, non-Open Source, kernel modules. Besides, oracle-asmlib is not needed for running ASM. The default Udev and Multipath work just fine. And originates from, and are being used in, some of the largest Linux clusters in the world, with PBs of cluster storage. Why RHEL and not OL (Oracle Linux)? RHEL5 is also a tad old. Why are you still on it? Old hardware? (am using OL5 on a couple of servers due to it still supporting old Mellanox HCAs).
>Why RHEL and not OL (Oracle Linux)?Some large organisations still chose RHEL instead of OL for servers that will be running only an Oracle Database.  I guess their logic is "We are running RHEL for the application servers, other database servers , etc etc ... so why choose OL and deviate from our 'standard'".  Not that I agree with this logic.  Given a choice, I'd use OL even if it is the lone database server among 100 other application servers.  Surely RHEL administrators can adminster OL servers !  Hemant K Chitale
In my opinion OL is better than RHEL. Has UEK support. Has better Infiniband support. Has Public Yum and is free when a support contract is not needed (no such option with RHEL itself). And a question in https://community.oracle.com/community/server_%26_storage_systems/linux/oracle_linux/content is likely to be seen and answered by an Oracle developer on the OL team. Also cheaper. One of the reasons we went from RHEL to OL some years ago was that the OL support contract for our 70 odd Linux servers was cheaper.
The solution for this problem has already been provided at How can I download the Oracle ASMLibs for Red Hat Enterprise Linux  5 ? and also in a post you created in yet another forum when will the oracleasm libs be availabe for kernel-2.6.18-410.el5.x86_64.rpm. You may have gotten away this time, but it is not a good idea to post your problem all over the place and hijack other people's threads, which is considered rude and not a welcomed practice.
People often don't know the difference or try to get free support for RHEL from OL, since RHEL is not a free product. Whether people use RHEL or OL may also depends on the hardware and what kernel versions are supported by the vendor. RHEL and OL also provides specific solutions that are not available in other distributions and require a support license, such as Red hat Gluster storage or Oracle Ksplice, for example. However to install and run Oracle database, there is usually very little point not to use Oracle Linux, and if I remember previous discussions correctly, even though Oracle Linux also ships with the so called RH compatible kernel, the Oracle Linux team only tests it's products on Oracle Linux.
It's true we only test Oracle products on Oracle Linux (because it's binary compatible to RHEL, so we can paper certify RHEL). However, we do test both the RHCK and UEK, because the RHCK ensures that it'll work on the equivalent RHEL version.

Nimble storage NCM

We are looking to use NIMBLE storage.We have the followingOracle 11GR1 on OEL 5.7 using ASM storageOracle  11GR2 on OEL 6.3 using ext3 file system  Will NCM be compatible with these versions of Linux and Oracle.
You would need to ask Nimble, not us.
Avi is right. I'll add there is a configuration Oracle has tested in conjunction with Cisco and Nimble that comes close: Oracle Linux 6 Update 5 x86_64 on Cisco Unified Computing System B200 M4 and Nimble Storage CS Series with Oracle Database 11g Release 2
Maybe it would be helpful to mention that the product is no longer called OEL, but OL for Oracle Linux.  OS 5.7 and 6.3 are old. There is probably no valid reason not to update to 5.11 or 6.9 - it will most likely increase support changes. 5.7 already included the Oracle UEK kernel, which is the default kernel since 5.6, however, If the storage is compatible with RHEL or CentOS, then it should also be compatible with Oracle Linux, which also includes the RHCK kernel.

Migrate RHEL to OEL.

Hi experts, we are planning to migrate our 2000 RHEL ( 5,6,7) physical and virtual server  to OEL.  I search this https://linux.oracle.com/switch/  Using this procedure, will upgrade the packages from ULN and also UEK kernel will be installed? what are precaution we have to follow before upgrade our RHEL to OEL? we have both oracel RAC and application servers  weblogic/websphere/MQ are there. ThanksVeera.
From what I know, each vendor provides some exclusive services and software. If you are using any exclusive services, you need to make sure an alternative exists, for example, ksplice versus kpatch. Generally speaking, OL is 100 % binary compatible and from the same source. OL ships with the Red Hat kernel included, but as of OL 5.6, the Oracle UEK kernel is the default. The UEK kernel is optimized for Oracle products, which will probably simplify support in your case, but keep in mind that UEK and RHCK do not necessarily have the same vendor support for special kernel drivers you may need depending on your hardware. I suggest you review some of the information available at https://linux.oracle.com/

Categories

Resources