Hi Dominic, et al: Remember me? I wasn't sure if I should post this to the list, cc the list, or just email you directly, since my next question is probably technically more general to linux than to TSL, but some of our process definitely seems TSL specific. I guess I'll cc the list for now, and If you or others let me know the preferred method on that, then I'll proceed accordingly.
Scenario: We've decided to go with TSL 2.2 on an HP server platform. Some DL385 and some DL145 boxes, Equallogic iSCSI SAN, and the Qlogic 4010-C HBAs We've decided to build an interim network first, and migrate our existing low availability network to the SAN using boot from SAN, which should give us the ability to recover very quickly from most hardware failures by simply booting to the failed server's volume on the SAN from a stand-by server. Once the interim network is stable, we'll start working toward the high availability network, probably in two stages. First an Active-Passive arrangement, and moving toward an Active/Active load balanced cluster scenario (more details when we get closer) We've reached the point where we need to compile the Qlogic HBA drivers into the TSL kernel. Our first steps are to install TSL on the selected hardware and compile the Qlogic drivers on that platform. We'll then create a dd floppy and do a fresh TSL install using the command line 'linux dd' option, and hope TSL will load the drivers at boot time and be able to do a boot from SAN. If that works, we'll want to compile the drivers directly into the TSL kernel. We'll then create a PXE/TFTP server on the LAN and plan to perform TSL installs on our servers, without hard drives, directly onto the SAN and configured to boot from SAN. Big problem here, is that I've never been exposed to compiling kernels in any type of custom fashion. Questions: 1. I'm uncertain if the TSL kernel is identical to the kernel we might download at say kernel.org, or if the TSL kernel has modifications requiring us to get the kernel from Trustix and work from there (which is what I'm guessing)? 2. Would someone be able to point me in the direction of a good how to, either on compiling the TSL kernel with custom drivers, or generally compiling the Linux kernel with custom drivers? I'm not quite sure how to approach that, and whether we can somehow add a module to the TSL install disk and have it compile at install time, or just how that all works. Any help is greatly appreciated, and I'll be happy to share any useful part of my experience once we get this all working. - Kenny On Nov 17, 2005, at 9:32 AM, Dominic Maraglia wrote: > Hello K, > > I work for a company named polyserve (polyserve.com) -- we are in > the business of making highly available clustered systems, based on > linux and also some windows products. I work in the linux group > and have much experience with setting up linux clusters. > > K Jepson wrote: > >> I want to build a highly available cluster using HP DLx servers on >> an Equallogic iSCSI Storage Area Network (SAN). >> >> One of my first concerns is regarding drivers. For instance, I >> currently plan to use the QLogic Host Bus Adapter (HBA) QLA4010- >> C. This adapter allows boot from SAN in addition to offloading >> the tcp/ ip workload from the cpu. >> > We also use qlogic hba's. For the most part the drivers are fine > -- in some cases the drivers have proven insufficient for our needs > and we have created some improvements in the qlogic drivers. These > fixes are to suit some of the special needs of our own software -- > I think the default qlogic drivers will be okay in most cases. > >> QLogic makes certain their driver stays built in to the RedHat >> kernel. Does that mean it would also likely automatically be >> supported in the TSL kernel? >> > It is easy to recompile the Qlogic drivers -- they provide the > source, and recompiling for a different platform is quite easy. > >> Second, I haven’t found any good posts in the archives re: >> building Highly Available Clusters (H-A Clusters) using Trustix. >> Could anyone point me in the right direction here? I’ll be happy >> to share my experience with this project for others to benefit >> from, but I’m hoping to find some initial information to get >> started. For instance, is there any type of H-A Clustering >> support built-in to TSL? If not, any ideas what is likely needed >> to make this work? >> > There are many, many variables that need to be considered first: > > How many server accessing the san? > > What file system will you use? Plain resiser? A clustered f/s like > GFS? > > What applications will you host on these severs -- H/A is not so > easy with some applications, but easier with others. > > Do you want your servers to act in "hot stand by" mode, or do you > want both servers performing workload at all times? > > Maybe you can provide more details about what you would like to > accomplish and I would be happy to provide you with much more detail. > > Cheers, > Dominic > _______________________________________________ tsl-discuss mailing list [email protected] http://lists.trustix.org/mailman/listinfo/tsl-discuss
