tnorman
asked on
Netward 5.1 - Replicas not syncing in timely manner
I have a four server Netware 5.1 setup. All servers have been patched with SP6e, and are running DS NLM 8.85.
There are four servers, server 1 through 4. SERVER1 is the Master replica, with a R/W replica on 2,3,4.
When I do a DSRepair, 'Report Sync status...' I get results like this (going with a current time of 8:45)
Replica Server1 9-07-2004 8:27:34
Replica Server2 9-07-2004 8:39:34
Replica Server3 9-07-2004 8:43:08
Replica Server4 9-07-2004 8:44:43
My biggest problem is when I create users, we have to wait 15-30 minutes for everything to sync, and then they can log in.
My biggest personal quandry is why the Master replica is always the oldest.
I would assume, with all servers connected so closely, there shouldn't be any lag (esp. with high-priority items like password changes and user creation). There are no communication problems between servers, and workstations can access all no problem. Time is reported in sync.
Any insight, thoughts, opinions welcome.
Thanks,
TN
There are four servers, server 1 through 4. SERVER1 is the Master replica, with a R/W replica on 2,3,4.
When I do a DSRepair, 'Report Sync status...' I get results like this (going with a current time of 8:45)
Replica Server1 9-07-2004 8:27:34
Replica Server2 9-07-2004 8:39:34
Replica Server3 9-07-2004 8:43:08
Replica Server4 9-07-2004 8:44:43
My biggest problem is when I create users, we have to wait 15-30 minutes for everything to sync, and then they can log in.
My biggest personal quandry is why the Master replica is always the oldest.
I would assume, with all servers connected so closely, there shouldn't be any lag (esp. with high-priority items like password changes and user creation). There are no communication problems between servers, and workstations can access all no problem. Time is reported in sync.
Any insight, thoughts, opinions welcome.
Thanks,
TN
ASKER
I could certainly remove a r/w off of one of the servers. Any 'forewarnings' about doing this?
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Forwarnings. Hmmm.
If the server is a busy server, don't do it during the business day unless your replica is small. Don't want to drag the server's performance down re-working its NDS database while the CEO is trying to arrange a golf date with his buddies. This is strictly a performance issue (I know a moron who removed a 2000+ object R/W replica in a 400+ server NDS tree from a server performing massive file service for over 800 client machines in the middle of the business day - the server's performance almost ground to a halt - NetWare v5.1 and an older NDS version, might have even been a late v7.x).
Do make sure the replicas, all of them, are healthy. Use DSREPAIR, get it down to 0 errors, start with the Master, then the other ones. 95% of minor NDS errors will have no effect in your situation, doing what you want to do; but a few runs of DSREPAIR are very cheap insurance.
I want to make certain that time really is in synch. I say this because I got burned early in my NDS career by not distinguishing between "Time Sycnhronization is Active" and "Time is Sychronized to the Network" when I entered "TIME" on the server console. The former statement merely means that the server has TIMESYNC.NLM loaded and that it is performing synchronization activities. The latter is what's *really* important - it means that the server is in agreement with the other servers as to what time it is. The latter is what you should see on all servers.
I would also make sure that the server that holds your Master replica is the Primary timeserver, and that the other ones are Secondary timeservers.
Removing an NDS replica should be fairly quick and painless in your environment. As I recall, that was accomplished using PARTMGR.EXE on a workstation logged in with admin-level privledges. Do NOT use NWCONFIG to "Remove Directory Services" from the server. :-)
If the server is a busy server, don't do it during the business day unless your replica is small. Don't want to drag the server's performance down re-working its NDS database while the CEO is trying to arrange a golf date with his buddies. This is strictly a performance issue (I know a moron who removed a 2000+ object R/W replica in a 400+ server NDS tree from a server performing massive file service for over 800 client machines in the middle of the business day - the server's performance almost ground to a halt - NetWare v5.1 and an older NDS version, might have even been a late v7.x).
Do make sure the replicas, all of them, are healthy. Use DSREPAIR, get it down to 0 errors, start with the Master, then the other ones. 95% of minor NDS errors will have no effect in your situation, doing what you want to do; but a few runs of DSREPAIR are very cheap insurance.
I want to make certain that time really is in synch. I say this because I got burned early in my NDS career by not distinguishing between "Time Sycnhronization is Active" and "Time is Sychronized to the Network" when I entered "TIME" on the server console. The former statement merely means that the server has TIMESYNC.NLM loaded and that it is performing synchronization activities. The latter is what's *really* important - it means that the server is in agreement with the other servers as to what time it is. The latter is what you should see on all servers.
I would also make sure that the server that holds your Master replica is the Primary timeserver, and that the other ones are Secondary timeservers.
Removing an NDS replica should be fairly quick and painless in your environment. As I recall, that was accomplished using PARTMGR.EXE on a workstation logged in with admin-level privledges. Do NOT use NWCONFIG to "Remove Directory Services" from the server. :-)
ASKER
I think we are on to something here. Of the four servers, this is what is happening:
SERVER 1 - OK
Server 2 & 3 - error, failed Replica in Skulk 698
Server 4 - Very odd:
Sync - [0000841b] <usrcat.wic.wic> [2002/03/28 16:19:45 ,2 ,1]
(this is scrolling/repeating about once a second on the screen)
('usrcat' is our Catalog for NDS login/wildcard searching.)
I am guessing I have some corruption somewhere here, and it is tying up the NDS update. The server that is having this repeating sync message is the one that is always behind in the sync.
SERVER 1 - OK
Server 2 & 3 - error, failed Replica in Skulk 698
Server 4 - Very odd:
Sync - [0000841b] <usrcat.wic.wic> [2002/03/28 16:19:45 ,2 ,1]
(this is scrolling/repeating about once a second on the screen)
('usrcat' is our Catalog for NDS login/wildcard searching.)
I am guessing I have some corruption somewhere here, and it is tying up the NDS update. The server that is having this repeating sync message is the one that is always behind in the sync.
Hmmm....yes, I think you're onto something. Definitely should not be seeing those messages constantly like that.
I confess to never having worked hands-on with NDS Catalogs, so I'm at a loss to suggest specific remedies. I would probably try deleting and re-creating the catalog (assuming there are no repair utilities to try first).
I confess to never having worked hands-on with NDS Catalogs, so I'm at a loss to suggest specific remedies. I would probably try deleting and re-creating the catalog (assuming there are no repair utilities to try first).
ASKER
I think I will close this question, and start off (at least) one new one with that specific error message going on. I will also post the message in Novell support forums.
Thanks for your help on this one.
TN
Thanks for your help on this one.
TN
NDS is a actual multi-master replication environment (AD claims it is, but its really master-slave), so the Master replica isn't a whole lot different than a R/W in most respects. The point I'm trying to make is there's no special reason he should be the first one updated.