Having trouble with NetWare? On July 25th Ron Nutter answered questions about those annoying little problems that can crop up using NetWare.If you couldn’t join us then, enjoy the transcript; and we hope to see you on our next live Guild Meeting. You can find a schedule of Guild Meetings in your weekly TechProGuild Notes TechMail, or on the Guild Meeting calendar.
Having trouble with NetWare?On July 25th Ron Nutter answered questions about those annoying little problems that can crop up using NetWare. If you couldn’t join us then, enjoy the transcript; and we hope to see you on our next live Guild Meeting. You can find a schedule of Guild Meetings in your weekly TechProGuild Notes TechMail, or on the Guild Meeting calendar.
Note: TechProGuild edits Guild Meeting transcripts for clarity.
Welcome to the Guild Meeting
MODERATOR: Tonight's topic will be about figuring out some of the most common problems with NetWare. We'll also give you some ideas about troubleshooting your NetWare servers. Our guest speaker tonight will be TechRepublic's frequent NetWare contributing editor Ron Nutter.
RON NUTTER: Thanks to everyone for stopping by.
RON NUTTER: Troubleshooting Novell is partly science and partly being methodical. One message that had led me down the stray path more than once is the good old message that mentions a page fault. The page fault message doesn’t really mean what it says. The message implies that you have a bad area in memory, but what the message means is that you have an NLM or other module that isn’t managing its memory correctly.
While keeping your server at the latest revs of particular modules is very important to having a trouble-free situation, it is a very good idea to keep at least one back rev of the modules that are being updated just in case a newer rev of an updated module causes a problem.
JCARLISLE: I guess the mix-up is most common with support packs (SP), right?
RON NUTTER: Support packs can be part of the problem but more likely just a symptom. For example, I have had a problem in the past when applying support packs on a server that uses NetWare for SAA and Hostprint. There is a common NLM (I think it is CSL.NLM) that Hostprint makes calls to that aren’t always supported in newer releases. That sometimes means a downreving of that one NLM to keep everybody happy.
TIM: I think SP8a for NW 4 goes into great detail about avoiding problems with NetWare for SAA.
RON NUTTER: Tim, I’m glad to hear that. I haven’t worked much with SAA in the past few months, but I have always been really careful around that one, long before development was taken over by IBM.
One thing that I will say about Novell is that with NW 4 and later, the ABEND.LOG can give you a good idea of the module that is having the problem. That comes not so much from looking at the module that it reports as being the problem but the modules that are listed in the stack close to where the "problem" module is listed.
TIM: The smartest thing to do with any NW SP is to read the text documentation because many potential problems are usually described in painful detail.
RON NUTTER: Tim, that’s true. I also usually wait until a NW SP has been out for at least a month before applying it. NW SP 4 was a good example of this; early releases introduced a TCP/IP routing problem that didn’t show up in all situations.
Examining memory management
JCARLISLE: Is memory management any better with 5.x than it was with 4.x?
RON NUTTER: Jcarlisle, I have had much better luck with 5 on memory, and my early tests with 5.1 indicate that the method they are using for memory management has been improved.
HAROLD966: Can you tell the difference between Novell and NT server memory management?
RON NUTTER: Harold966, I can tell a difference. I can let a NetWare server run for several months without having to restart it, whereas I can barely let my Exchange server go for much over a month without having to restart it to get the memory pools back to where they need to be. That is with NT 4, I will say; I’m still doing some testing with W2K.
Analyzing the Abend Analysis Site
TIM: Speaking of ABEND.LOG files, has anyone had much luck with Novell's so-called "Abend Analysis Site"?
RON NUTTER: Tim, I haven’t had much of a reason to use it. The ABEND.LOG or messages directly from the offending NLM itself have usually pointed me to the problem module.
One tool that I would suggest everyone take a look at is the Alexander SPK (Server Protection Kit). I have worked with an earlier version of this tool and was very impressed by what it did. What really got my attention was the capturing of screens present on the server at the time of the abend so you could see all the different screens and messages to see if the problem was due to more than one problem being present at the time.
MATTHEW.HILL: Is that the kit that was advertised for ages on Novell’s site? I love how they make a kit to stop it crashing. Why not build this technology into the original product we paid for?
RON NUTTER: Matthew.hill, this is a product that was created by a company external to Novell. Alexander enjoys an unusually close relationship with developers within Novell. The secret to how Alexander SPK works is that it uses a hook present in the operating system to get the information before the server comes to a halt. This hook was intended to be used by a similar product that Novell had planned to release but didn’t have the resources at the time to peruse.
In those cases when ABEND.LOG and other resources don’t point you in the right direction, I try to start with the basics. The basics being, take out what you last changed or start unloading modules one at a time to see when the error stops occurring. I had to take this last step, myself, back in the fall when my company was moving into a new building.
TIM: You know this can get pretty tedious when you have servers loading 100+ NLMs routinely.
RON NUTTER: Tim, you are right, but sometimes you may have no other choice. We moved 100 users and 5 servers in about 36 hours and were back in business. The problem was that we broke a few rules during the move: We installed a new server and upgraded one of the other ones. The problem on the primary server (NW 4.11) ended up being that we had placed the drives from an older level of hardware into a brand new system, and the network management modules were getting confused with the new hardware that they were seeing. Once we upgraded to the newer network management NLMs, the problem went away.
MATTHEW.HILL: Is this all that troubleshooting comes down to? When I was new to Novell, I assumed that the more experienced people would be able to do some advanced troubleshooting “things,” but it appears they're just the same basic things: What happened before the crash (then reverse it). From my experience, the ABEND.LOG can often be as helpful as a general protection fault.
RON NUTTER: Matthew.hill, there are some more advanced things that you can try, such as going into the NetWare debugger, but in most cases, staying with the basics usually gets me to the solution the quickest. The real secret is in knowing your server and how it works/responds on a daily basis. Contrary to what some people may say, machines can have a personality.
TIM: The SPK 2000 will have a feature that sends the EDNA crash file or something like it to Novell's Abend Analysis Site and receive a reply from the site right at the server's console. Internet access is a given, but if Novell can pull off the Abend Analysis Site with a successful launch this fall, we may finally be on our way to removing abends as the bane of our NetWare existence.
MATTHEW.HILL: Tim, how will the SPK2000 be able to send the abends for analysis if the servers are out of RAM or the NIC has a fault?
RON NUTTER: Matthew.hill, that is the one point of vulnerability in that feature. That is why I usually create a large DOS partition (usually 300 MB or larger) just in case I have to set the switches to allow the server to crash hard so that a memory dump file can be created.
TIM: Depending on the type of abend (soft or hard), the SPK can put the EDNA crash file out on the DOS partition and copy it back to the SYS volume on the restart if need be. And then I assume it would be able to submit it to the Abend Analysis Site.
TIM: There is a lot of useless stuff in a core dump. The SPK reduces everything of interest to about 25 MB when it builds the EDNA crash file.
NetWare 4x memory utilization
76327.711: I have a basic question about memory utilization on NetWare 4.x (regular and SBS) servers. We have several servers with 128-MB RAM that report as follows (with no users logged in) under Monitor/Memory Utilization: 4 kb Cache Pages: 1438; Cache Page Blocks: 253; Percent in Use: 71%; Percent Free: 29%; Memory Blocks in Use: 20,750; Memory Blocks Free: 807. Shouldn't we have more "free" everything at this point?
RON NUTTER: 76327.711, I would think you should have more available memory. What size hard drive and type of interface are you using?
76327.711: Hard drives are 6 GB and 8 GB, both EIDE.
RON NUTTER: 76327.711, what other NLMs (beyond those required for disk controller and network card) do you have loaded? Have you tried using the Purge command with the /all option running from the root of each volume? Since NW keeps track of deleted files that can tie up memory as well.
76327.711: All "standard" NLMs are being loaded: those lined up for loading when the server was first set up. I'll run Purge /all and see if things improve. Any other suggestions?
TIM: Another avenue to try with NW 5 troubleshooting is to run the offending NLMs in extended memory, if possible, where they can't take down kernel memory. Some caveats on what can and can't be run in extended memory, but certainly worth a try. Performance hit will be about five to 10 percent. Actually, you can run all of GroupWise in extended (ring 3) memory.
RON NUTTER: 76327.711, I know that NW 4 uses its memory better than earlier versions, but with the amount of disk space you are mounting, I would think that more memory might be needed. My old rule of thumb is to have 16 MB of memory for every gig of mountable hard drive space. While this usually is more memory than you can get by with, it will keep you running with plenty of caching memory to spare.
By my calculations, this would mean you might need upward to 256 MB of RAM in the server. While I have been criticized in the past for this type of calculation, I have found that the Novell calculations have been somewhat "conservative." That is not a criticism of Novell, but just a reflection of my day-to-day experiences with the product.
76327.711: We've had no problems running these servers; it just seems that there should be more "free" available. Using your formula, 128-MB RAM is "correct" for an 8-GB hard drive.
RON NUTTER: 76327.711, your message indicates that you were mounting a 6-gig and an 8-gig drive at the same time. Correct?
76327.711: Sorry, these are independent servers with single HDDs each: either 6 GB or 8 GB.
TIM: I don't build a NetWare 5 server today with less than 256 MB of RAM, and I recommend 512 MB.
RON NUTTER: Tim, you follow the same rule I do. Memory has been cheap enough lately to make that one easy to get by my management folks. It helps when the IS director understands this concept and stands behind you when you’re making the recommendation.
76327.711: Tim, when you build a 256-MB server how much free memory is available when the server initially loads?
TIM: I always shoot for 60 percent or higher in file cache buffers.
RON NUTTER: Tim, that’s exactly the figure I like to use. With less active servers, I have been able to run with just under 50 percent comfortably, but I try to keep the more active servers over 60 percent. When I start hitting 60 percent, I usually look to add more memory or offload some of the work to another server.
TIM: Of course, with NW 5 you can make use of swap files too, which can ease the pressure on RAM.
RON NUTTER: Tim, when using swap files in that situation, I would prefer to put the swap file on a separate volume with no other files present so you are not fighting for resources on the same drive when you have to start the swapping process.
TIM: LRU sitting time is always a good indicator of whether or not the server has enough RAM. An LRU value of less than 15 minutes is a serious problem that should be fixed with more RAM.
Swap files and FastCache proxy volumes should go on separate and/or multiple volumes.
MATTHEW.HILL: I live in New Zealand, and the general consensus here is that Novell is dying out and will most likely sell its eDirectory/NDS to another company like Sun. Are the feelings in the U.S. the same?
RON NUTTER: Matthew.hill, that is a case of Microsoft marketing. Novell realized several months ago that the server OS market wasn’t where things were probably going to head in the long term. I was fortunate to have a private talk with Drew Major one evening during Brainshare U.S. and got an idea of where he sees things heading. This is another reason that Novell is heading to the "services" market with offerings such as their recent "onDemand" product. This product is targeted at the ASP community and allows the ability to "rent" software for a use-on-demand situation, subject to software license agreements. This appears to target small to medium sized businesses but is definitely an area that needs to be addressed. Novell is also strongly going after the "directory" market. They get the message from the customer base that wants to get all their products working together regardless of the OS. I don’t think Microsoft sees that same picture. Novell did try to downplay the future of NetWare at Brainshare but quickly changed its tune. NetWare 6 and NetWare 7 are on the boards.
76327.711: The server scene would have a big hole in it without NetWare.
MATTHEW.HILL: So you do see the decline of the NetWare server market? Will this support be going west, too? I read an interesting article that was called "How do you sell off a company with a billon dollars in the bank?"
RON NUTTER: Matthew.hill, I don’t see a decline of the server OS market but see new opportunities where they can provide goods to meet a need.
I can remember in the networks that I first put in that people were happy to have a large drive to put their files on. Now that feeling is "I have a large drive in my workstation. What else can the server/network do for me that I can’t do for myself?"
76327.711: The "traditional" server cannot be abandoned. Not every company is Fortune 100 in size and in a position to take advantage of, or need the power of, eDirectory to run three-user accounting software, four- to five-user point-of-sale systems, and so on.
RON NUTTER: 76327.711, I don’t think you will see Novell abandoning the traditional market.
TIM: NetWare exists today as a platform for network and Internet services, and NDS is the glue.
MATTHEW.HILL: Exactly. What can you leverage off of Novell that you can't get with WinNT in a File & Print environment without the integration issues? The only reason my company is recommending NetWare at present is because of desktop management with Zenworks.
76327.711: Novell has one BIG advantage over NT: It keeps on running. There is no comparing the stability of NetWare over NT.
RON NUTTER: I think the message is “Don’t count Novell out just yet.” Each year I attend Brainshare, I see more things that Eric Schmidt has gotten the company to do and new directions they will be heading.
TIM: Dr. Schmidt could use a good right-hand man or woman at Novell.
RON NUTTER: Tim, I think Dr. Schmidt is doing just that. Look at the recent senior management changes. He also believes strongly in the town hall meeting concept to get ideas to market sooner.
RON NUTTER: 76327.711: That is very true. Now the trick is to get senior management and the decision makers in companies to understand that concept and not listen to marketing hype.
76327.711: I like Eric Schmidt. I just hope that he and his team can weather the MS storm and the public perception of the server competition between the two companies.
TIM: Actually, I hope Dr. Schmidt can weather the Novell board of directors. Apparently, there is someone on the Novell board who wants him ousted.
TIM: Novell will release an OEM NetWare NAS product soon. Like ICS, it will be branded by the manufacturer with NetWare under the hood.
RON NUTTER: Tim, this is addressing a market that Novell saw coming, almost a plug-and-play type of situation. For those who weren’t at last year’s Brainshare, they had a ten-year-old kid configure an ICS box in a little less than 20 minutes. Granted, it was Drew's son, but he did it faster than the others they had onstage.
Internet Caching Service
76327.711: What is "ICS?" Internet Connected Server?
RON NUTTER: ICS stands for Internet Caching Service. ICS is used for those companies that want to make their Web content more available on the Internet and reduce some of the load on the Web servers that keep sending out the same information again and again.
Thanks for coming
MODERATOR: Thanks to everyone for their participation and thanks again, Ron, for a very informative presentation
Our Guild Meetings feature top-flight professionals leading discussions on interesting and valuable IT issues. You can find a schedule of Guild Meetings in your weekly TechProGuild Notes TechMail, or on the Guild Meeting calendar.