|
Home > Archive > CNE > June 2001 > Replica Placement
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
|
|
| Freddy 2001-06-21, 8:43 pm |
| Replica placement is about to drive me nuts...as though that were a long drive...
Does anyone have any memory tricks that helped them get it straight? Any help would be appreciated. | |
| dheinsdorf 2001-06-22, 9:05 pm |
| quote:
Originally posted by Freddy
Replica placement is about to drive me nuts...as though that were a long drive...
Does anyone have any memory tricks that helped them get it straight? Any help would be appreciated.
No but stay tuned. I am going to hit the 575 books this weekend. I had a really good instructor for the course, one of the best in Phoenix, and I took massive notes, so I will pass it along 
I got a week and change to take the test before my Vue voucher expires so talk to ya soon.
Doug | |
| dheinsdorf 2001-06-25, 9:13 pm |
| quote:
Originally posted by Freddy
Replica placement is about to drive me nuts...as though that were a long drive...
Does anyone have any memory tricks that helped them get it straight? Any help would be appreciated.
Freddy,
I hit section 3 of the student manual today. When I said my instructor of this class was good and that I took massive notes, well she was good because she related to the text so that I understood it and didnt mark the heck out of the manual.
What about it do you feel your not "realizing" at this point? Like DS versus file system rights amongst many other things, I often dont get some things till they "snap" into learnedness. Besides the more granular rules involved with replica placement, I get this. But I dont have any tricks - like 5.x admin's BMDOSTP:
Backup
Manage
Document
Organize
Setup
Tune
Protect
I made up that one for 654.
Anyhow, do you have a 575 student manual? If so are there any particular theories you havnt snapped on yet? The manual is pretty killer as I am going back over it.
Doug | |
| chunder 2001-06-25, 11:44 pm |
| the 4 types:
Master, Read/Write, Read-Only, Subordinate Reference.
Master and R/W can make changes to the Directory.
Read-Only are lame.. really. well, ok, they exist, know that. they do nothing but auth and pass NDS changes to M or R/W (meaning they don't do changes).
Sub-Refs are merely pointers to where the actual replicas are. you don't create and you don't delete (don't means can't). quote:
Subordinate reference replicas are only created on servers that hold a replica of a parent partition but no replicas of its child partitions.
think "where's my baby!?!?!" mother shreeks histerically! "over here! over here!" replies the SR. basically it tells NDS how to get to the child partitions.
or in other words:
if a server has a replica of a partition, and that partition has a child but the server doesn't have a replica of that child, then the server will also have a SR of the child.
replica placement is fun. you will want to get this chart stuck in your brain. (http://www.novell.com/documentation...a/h3he2ozs.html)
and this is good too: http://www.novell.com/documentation...a/fbaecheh.html
by default we know that the first server installed has a master replica of ROOT. the next two servers installed into the tree get R/Ws. that's it. them's the default.
always have 2 or 3 replicas of each partition. i think for the test, know 3. have no more than 10 replicas on a server. i have actually heard of companies so large (nds-wise) that they have servers just for NDS Replica Servers. WOW!
another: http://www.novell.com/documentation...a/h3yj5jw4.html
all kindsa good links: http://search.novell.com/NSearch/Se...html&numhits=20
and don't worry if you see a lot of 4.11 and 4.2 references on some of those links -- it's all NDS baby!
there's a PDF called taoenu.book in that last link.. it's a 6 meg pdf but full o' good reading.
HTH | |
| dheinsdorf 2001-06-26, 6:53 am |
| Good stuff Chunder, Freddy might focus on the one you listed at:
http://www.novell.com/documentation...a/h3yj5jw4.html
I think that one and its graphic do a better job than the student manual at explaining it.
Heres the short of how I remember it from a disection versus design standpoint. this assumes someone else already designed the replication strategy:
Reminder Rule.
-Find parent partition (use graphic to identify parents and children). Find "the" server holding M,RW or RO of that partition.
-Does "the" server have parent or child partitions?
-If "the" server has parents and/or children, but doesnt contain M/RW/RO of those parents and/or children, place SR on "the" server.
That assumes for a chart with the SR's not filled in. It also helps me to draw up a mock tree with geographic or function based boundaries, a simple pyramid. place some servers and draw some partitions. Fill in the chart as follows:
-Partitions on top
-Servers on side
-Fill in M masters of each partition for a particular server
-Fill in RW read write's based on 2 copies of "partition" root (not to be confused with [Root] partition, which also gets included).
-Then add in SR's based on Reminder Rule.
If this doesnt seem right, anyone please let me know.
Doug | |
| Freddy 2001-06-26, 8:50 am |
| Doug and Chunder,
You guys nailed it right on. The problem I was having with placement and remembering which way to search the chart, it just snapped into place with the mother searchihg for her children analogy. Thanks for the links, they are very helpful.
I am using the Clarke book for CNE study. It is written pretty well and I am grasping the concepts without much trouble, it was just the placement that was driving me crazy.
I scheduled my exam for Saturday 6/30, I'm going to take the 634 and try my hand against the adaptive test. It scares me half to death but then that will increase the excitement a bit. I'm getting bored with the big long form tests. Thanks again guys, this should help a lot of people that are struggling with NDS. | |
| dheinsdorf 2001-06-26, 12:20 pm |
| I screwed up in my "thinking" of how to easily remember the placement of SR's, so I am going to try again, this is the one I will use to get through this:
After the M's and RW's are placed where you want them in the replica table:
-From tree structure graphic, jot down parents and children for partition boundaries.
-Start from the top of the Server side of the chart, left to right.
-If the server does not have a copy of the parent partition, no SR's required. X out open spaces to right for each partition.
-If server has a copy of the parent partition = need SR of all the parent partition's children. Continuing from the left to right across that server line, ask in each space if that partition is a child (of the parent of the child partition you are referencing for this particular server).
I think that is better and more correct. I got my Manager, an old school 3/4/5 CNE to help me out.
Not a bad idea to take the adaptive. I may do it myself 6/29, but the form test has a nice thrill of the hunt feeling to it. | |
| Freddy 2001-06-26, 1:33 pm |
| Doug,
I like the methodology of this way better. Putting an X in the box definitely tells you where you are at while you are analyzing a chart. Systematic analysis really helps to avoid errors. It could avoid confusion when you are looking back to see what you have.
Form tests have a certain appeal to me, I get to see more questions and test my knowledge against a wider variety of topics than in an adaptive test. The adaptive kind of appeals to me because each time you answer correctly you get a more difficult question to test yourself against. It sounds kind of fun in a way. I have to try it once to see what it is like to challenge one. Curiosity is killing the cat...
Good luck on your test. If your testing Friday, let me know what test you take and what you think of it. I'll do the same on Saturday. | |
| dheinsdorf 2001-06-26, 2:14 pm |
| quote:
Originally posted by Freddy
Doug,
I like the methodology of this way better. Putting an X in the box definitely tells you where you are at while you are analyzing a chart. Systematic analysis really helps to avoid errors. It could avoid confusion when you are looking back to see what you have.
Form tests have a certain appeal to me, I get to see more questions and test my knowledge against a wider variety of topics than in an adaptive test. The adaptive kind of appeals to me because each time you answer correctly you get a more difficult question to test yourself against. It sounds kind of fun in a way. I have to try it once to see what it is like to challenge one. Curiosity is killing the cat...
Good luck on your test. If your testing Friday, let me know what test you take and what you think of it. I'll do the same on Saturday.
Right arm man! I meant right on! (get it, right arm - your avatar)
Great CNE group session for sure (to all involved), got me thinking and I gained from it. You should have seen my Manager grabbing a root trying to remember SR's this morning. CNE training is good stuff, took him only a few moments. Novell trains us tuff. | |
| chunder 2001-06-26, 2:23 pm |
|
glad i could have helped some... | |
| Freddy 2001-06-26, 3:20 pm |
| Doug and Chunder,
Thanks for all of the great info you guys shared in this thread. I think this will be a thread we can link people to in the future when they are agonizing their way through NDS.
Your right, Novell trains us tough, but would we really want it any other way...
That right arm is doing a very important job, it is treating a laptop with all of the respect it deserves... | |
| dheinsdorf 2001-06-27, 4:56 pm |
| More replica placement memory tricks:
- First thing, on the Replica table. Go vertically down the [Root] partition column.
Any server that has a copy of the [Root] partition should have no empty spaces across that servers horizontal. If a server has a copy of [Root], plug in SR's all the way across.
- Go vertically down each partition, circle the M. If the partition (partition root) doesnt have an M it's wrong. If it has more than one M it's wrong.
- Go vertically down each partition, make sure there is also two R/W's.
- Now look at the list of answers for the test question. It may be quicker to reference each possible question against the Replica Table and/or NDS Tree than to further continue solving/resolving the whole Replica Table (if its big).
Reason why I say this is under STS questions, there are more things incorrect on the Replica Table than exist as answers, so I was wasting time on like "an SR that should'nt be there" when it was'nt a possible answer. | |
| dheinsdorf 2001-06-27, 11:16 pm |
| Wooah! I screwed up. The reference below to checking off all SR's from the [Root] partition horizontally for each server is incorrect. For whatever reason I was thinking that the Parent flows down the Tree and crosses partition boundaries, in fact the parent is the partition immediately above the root of another partition. I failed to notice the extra SR's = not good.
I appologise for the incorrect theory.
Doug
quote:
Originally posted by dheinsdorf
More replica placement memory tricks:
- First thing, on the Replica table. Go vertically down the [Root] partition column.
Any server that has a copy of the [Root] partition should have no empty spaces across that servers horizontal. If a server has a copy of [Root], plug in SR's all the way across.
- Go vertically down each partition, circle the M. If the partition (partition root) doesnt have an M it's wrong. If it has more than one M it's wrong.
- Go vertically down each partition, make sure there is also two R/W's.
- Now look at the list of answers for the test question. It may be quicker to reference each possible question against the Replica Table and/or NDS Tree than to further continue solving/resolving the whole Replica Table (if its big).
Reason why I say this is under STS questions, there are more things incorrect on the Replica Table than exist as answers, so I was wasting time on like "an SR that should'nt be there" when it was'nt a possible answer.
| |
| Freddy 2001-06-28, 9:22 am |
| Good job Doug, if you hit another partition in the chain, that stops the flow. But you are right in following the chart to the next partition boundry which should all be populated with SR's or R/W's. Maybe an R/O thrown in for fun.
One master per partion, that would definitely create a problem if there are more than that. I found in following the charts in EE55 that the problems are usually the S/Rs or no master of a partion stored anywhere. Quickly following out the charts like you recommend would save valuable time on test day. |
|
|
|
|