toasters September 2003

toasters@lists.teaparty.net

72 participants
70 discussions

Dump performance
by Duncan Greenwood 26 Sep '03

26 Sep '03

Boring I know, but I'd be interested to learn what kind of backup performance people are getting for Windows/CIFS shares. I'm particularly thinking of home and shared directories with millions of small files - the usual junk that accumulates on these types of shares with hundreds of users. I've seen amazing performance from dumps of Oracle datafiles, in the order of 90MB/s to three fibre-attached LTO-1s from an F840, but I know many small files will reduce thruput significantly. The question is how much and what would you do differently ? I'd be grateful for any numbers people could share. We'll probably use NDMP but I guess simple dumps would give a similar performance. Thks in advance. D # --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.521 / Virus Database: 319 - Release Date: 23/09/2003

1 0

Alternatives?
by Tuc 26 Sep '03

26 Sep '03

Hi, We currently have a customer with an F760 and 6 shelves of 36's, and an F720 with 6 shelves of 18's. We've had problems with both that of course should never have happened. About the only thing we came up with is another F760 with 6 shelves of 72's, vol copy to it, and then having the customer FTP to both to keep things in sync. This solution is a bit pricey for the customer. (About $47K) Has anyone looked at any of the IDE based solutions to replace a F760 class filer? We do about 4500 to 5K NFSops. Anyone had any other ideas? Thanks, Tuc/TTSG Internet Services, Inc.

3 2

C4.log?
by Gifford, Josh 26 Sep '03

26 Sep '03

Hello Toasters, In troubleshooting latency issues between our main site and a remote location, we have noticed a lot of traffic going to our filer and erroring out. Each machine in the remote location is trying to get to a file called \%username%\c4.log on a filer in our main site, but is unable to find it. Between all 20 users in this site, these events are occurring every minute or two. This traffic is coming from machines running Windows XP and 2000, some of the machines are in our active directory domain, some are in our NT 4.0 domain. I can't even find a single application outside of MS Office that they are all using. An example of the packet trace is provided below. Does anyone know what the c4.log is? Josh J. Gifford MCP Network Systems Administrator Siemens PTD 7000 Siemens Rd Wendell, NC 27591 Office 919.365.2806 Fax 919.365.1080 josh.gifford(a)siemens.com <mailto:josh.gifford@siemens.com> SMB command rejected Object not found Packet Info Flags: 0x00 Status: 0x01 Packet Length: 168 Timestamp: 13:33:15.081093 09/11/2003 Ethernet Header Destination: 00:05:5E:E3:98:40 Source: 00:0A:B7:47:92:F3 Protocol Type: 0x0800 IP IP Header - Internet Protocol Datagram Version: 4 Header Length: 5 (20 bytes) Type of Service: %00000000 000. .... Precedence: Routine ...0 .... Normal Delay .... 0... Normal Throughput .... .0.. Normal Reliability .... ..0. ECT bit - transport protocol will ignore the CE bit .... ...0 CE bit - no congestion Total Length: 150 Identifier: 56937 Fragmentation Flags: %010 0.. Reserved .1. Do Not Fragment ..0 Last Fragment Fragment Offset: 0 (0 bytes) Time To Live: 128 Protocol: 6 TCP - Transmission Control Protocol Header Checksum: 0x4ED9 Source IP Address: 161.134.199.208 Dest. IP Address: 161.134.194.65 No IP Options TCP - Transport Control Protocol Source Port: 1060 polestar Destination Port: 139 netbios-ssn Sequence Number: 266124745 Ack Number: 2144399882 Offset: 5 (20 bytes) Reserved: %000000 Flags: %011000 0. .... (No Urgent pointer) .1 .... Ack .. 1... Push .. .0.. (No Reset) .. ..0. (No SYN) .. ...0 (No FIN) Window: 16743 Checksum: 0x7ADF Urgent Pointer: 0 No TCP Options CIFS Header Reserved: 0x0000 Length: 106 SMB - Server Message Block Protocol ID: SMB Command Code: 50 Transaction2 - Function, Byte In/Out NT Status: 0x00000000 STATUS_SUCCESS Flags: 0x18 .... ...0 Does Not support Lock and Read/Write and Unlock .... ..0. Reserved (Must be zero) .... .0.. Reserved (Must be zero) .... 1... Pathnames are without case ...1 .... Pathnames are already in canonicalized format ..0. .... Do not request opportunistic lock .0.. .... Do not notify consumer on any modify action 0... .... Request Flags2: 0xC807 .... .... .... ...1 Application understands long file names .... .... .... ..1. Application understands extended attributes .... .... .... .1.. Use message authentication .... .... .... 0... Reserved for future use .... .... .0.. .... Is not long name .... 1... .... .... Client aware of extended security ...0 .... .... .... Application is NOT DFS Capable ..0. .... .... .... Application does NOT do Paging I/O .1.. .... .... .... Application understands NT Status Codes 1... .... .... .... Application understands Unicode Strings Reserved: ............ 00 00 00 00 00 00 00 00 00 00 00 00 Tree ID (TID): 0x0048 Process ID (PID): 0x0490 User ID (UID): 0x0800 Multiplex ID (MID): 0xEC43 SMB Transaction2 - Function, Byte In/Out Request Word count: 15 Total Param Bytes: 38 Total Data Bytes: 0 Param Bytes To Recv: 2 Data Bytes To Recv: 40 Setup Bytes To Recv: 0 Reserved: 0x00 Flags: 0x0000 Timeout (millisec.): 0 Reserved: 0x0000 Params This Buffer: 38 Params Bytes Offset: 68 Data This Buffer: 0 Data Bytes Offset: 0 Setup Word Count: 1 Reserved: 0x00 Setup Words: 0x0500 Byte Count: 41 Padding: 0x00000000 Trans2 Query Path Information: Get File Attributes Given Path Information Level: 257 Query File Basic Info Reserved 0 Must be zero Filename: \markspa\C4.LOG FCS - Frame Check Sequence FCS (Calculated): 0xE189B91C

1 0

Any news?
by John Enger 24 Sep '03

24 Sep '03

A while ago we had a debate on lisencing of NetApp software for non commercial/hobbyist use. We had a very promesing conclusion that the lisence system needed reviewing. I was wondering if there is any news regarding this? John

2 1

RE: Spares Kits
by Borders, Rich 24 Sep '03

24 Sep '03

Wanna talk to a sales person? :) Richard D Borders Filer Escalations Engineer RTP, North Carolina USA - Network Appliance, Inc. Email: rborders(a)netapp.com Phone: (919) 993-5866 > -----Original Message----- > From: Chuck Tomasi [mailto:chuck.tomasi@plexus.com] > Sent: Wednesday, September 24, 2003 4:49 PM > To: toasters(a)mathworks.com > Subject: Spares Kits > > I've got a few sites that need spares kits for their F720s. > Any clue where I can pick up such items? (CPU and FC8/FC9 > shelf spares) > > --Chuck > >

1 0

Spares Kits
by Chuck Tomasi 24 Sep '03

24 Sep '03

I've got a few sites that need spares kits for their F720s. Any clue where I can pick up such items? (CPU and FC8/FC9 shelf spares) --Chuck

1 0

very slow download ...
by Frank Bonnet 24 Sep '03

24 Sep '03

Hi I am upgrading a F87 to 6.4.2 and it seems very slooooow to perform the "download" command ... ( not finished yet after 20 minutes ...) any infos/ideas welcome ! thanks -- Frank Bonnet

2 1

How many spares? Best way to use extras?
by pdunkin＠lucent.com 24 Sep '03

24 Sep '03

Our F760 has 6 shelves and 42 36GB drives (running Data OnTAP 6.3.3 if that matters). Four volumes have been configured, with one RAID group in each volume; two have eight disks each, one has seven, and one that is mostly inactive archived stuff has four disks. That leaves fifteen (count 'em, 15) spares. All the volumes except the archive are at or near 90% of capacity, a point at which I understand performance starts to plummet, and the storage needs of the users of the three active volumes continue to increase. What is the optimal way to make use of the oversized pool of spares? Thoughts I've had: - Create a new volume, put new projects there, and maybe move some projects over from existing volumes, so everyone who needs it has room to grow. - Add disks to the existing volumes gradually as needed, to keep capacity under 90%. If I do this, is it better to add new disks one at a time or several at once? Is going over 90% really a problem, or is that just unfounded rumor? Other suggestions are welcome. Also, how many spares would it be appropriate to keep as spares? I've been told that one per shelf would be enough, but some postings in the archives indicate that that may be on the generous side. Thanks! -- Patricia Dunkin Lucent Technologies pdunkin(a)lucent.com 600 Mountain Avenue 3C-306C Phone: 908-582-5843 Murray Hill, NJ 07974-0636 Fax: 908-582-3662 Pager: 888-371-8506 Mailto: 8883718506(a)skytel.net

5 4

RE: How many spares? Best way to use extras?
by Hill, Aaron 24 Sep '03

24 Sep '03

Hey, toasters has been quiet lately, good to see some solid technical discussion on the agenda! The 90% issue. I have only ever seen it degrade performance from an application/user perpective when the volume is holding database device files and being pounded. For flat files that don't change much, I wouldn't stress. If you have large snapshots for say, snapmirror, then I wouldn't go above 80% just in case. Adding disk? Without sounding like a Netapp salesman, "the more the better". Never add one disk at a time. The rule is 3 minimum to provide better striping. Add one and you could see a significant performance hit as that disk thrashes away. NOW has more details on this. From a pure technical design and efficiency perspective, I like Jim's suggestion. If you can afford the disk, keep vol0 small and sacred, if you can't, then there is no reason why not to use the space. The major caveat is the full volume backups and restores. I would do some careful calculations based on what I know about my backup solution. If your daily data changes are large and your backup windows small, you will have some backup problems. Even though double disk failure is extremely rare as Jim points out, if you have SLA's that specify a strict full restore window for that volume AND the volume DOES die ............. I hope your restores are fast. Get your ducks lined up before you shoot with this one. I don't think backing out of the 1 volume will be as easy as going into it. Aaron -----Original Message----- From: Jim Harm [mailto:jharm@llnl.gov] Sent: Wednesday, 24 September 2003 3:03 AM To: pdunkin(a)lucent.com; toasters(a)mathworks.com Subject: Re: How many spares? Best way to use extras? Let's start a warm thread. My two cents is to: 1. build a ten disk raid from the 16 spares and set raid size to 9+1 2. migrate one of your eight disk raid/volumes to it as a qtree 3. destroy unused eight disk/volume 4. build another ten disk raid from left overs and previous volume 5. add this raid to the previous ten disk raid/volume to make big volume 6. migrate the next eight disk raid/volume to big volume as another qtree 7. continue to destroy old raids and migrate old to new volume as qtrees 8. when you are done you should have one big volume of four 10 disk raids and two hot spare disks. I've had two disks fail on one system once in 4+ years and that was during the disk firmware upgrade that only temporarily failed them (had to power cycle). We have had several TB of data on NetApp filers at this sight over that period. Use qtrees for all your exports to control space usage instead of building several volumes/raid strategies. It's simpler and more flexible. You can reapportion the space with little pain and no raid or volume reconfigures just by changing quotas. Some may say "Whoa, you have to have a separate root volume!" I say "Baloney!. The only time I came close to losing a root volume was in the infancy of NetApp; we did have to reboot and use a different volume as root volume (I had copied the /etc to the non-root volume) because of a problem with the wack program, but even then we recovered the volume. Another concern might be that "We're afraid the root volume will get full!" which you can easily avoid by judicious and simple quotas. The only other thing I can think of is the time it takes to back up and restore by volume size. At 11:56 AM -0400 9/23/03, Patricia A. Dunkin wrote: >Our F760 has 6 shelves and 42 36GB drives (running Data OnTAP >6.3.3 if that matters). Four volumes have been configured, with >one RAID group in each volume; two have eight disks each, one has >seven, and one that is mostly inactive archived stuff has four >disks. That leaves fifteen (count 'em, 15) spares. > >All the volumes except the archive are at or near 90% of >capacity, a point at which I understand performance starts to >plummet, and the storage needs of the users of the three active >volumes continue to increase. > >What is the optimal way to make use of the oversized pool of >spares? Thoughts I've had: > >- Create a new volume, put new projects there, and maybe move > some projects over from existing volumes, so everyone who > needs it has room to grow. > >- Add disks to the existing volumes gradually as needed, to keep > capacity under 90%. If I do this, is it better to add new > disks one at a time or several at once? Is going over 90% > really a problem, or is that just unfounded rumor? > >Other suggestions are welcome. Also, how many spares would it be >appropriate to keep as spares? I've been told that one per shelf >would be enough, but some postings in the archives indicate that >that may be on the generous side. > >Thanks! >-- >Patricia Dunkin Lucent Technologies >pdunkin(a)lucent.com 600 Mountain Avenue 3C-306C >Phone: 908-582-5843 Murray Hill, NJ 07974-0636 >Fax: 908-582-3662 Pager: 888-371-8506 Mailto: 8883718506(a)skytel.net -- }}}===============>> LLNL James E. Harm (Jim); jharm(a)llnl.gov System Administrator, ICCD Clusters (925) 422-4018 Page: 423-7705x57152 ************** IMPORTANT MESSAGE ************** This e-mail message is intended only for the addressee(s) and contains information which may be confidential. If you are not the intended recipient please advise the sender by return email, do not use or disclose the contents, and delete the message and any attachments from your system. Unless specifically indicated, this email does not constitute formal advice or commitment by the sender or the Commonwealth Bank of Australia (ABN 48 123 123 124) or its subsidiaries. **************************************************

1 0

non root volume etc was Re: How many spares? Best way to use extras?
by Jim Harm 23 Sep '03

23 Sep '03

When I backed up the root /etc directory from root volume to non-root volume, I did use rsync. About 2 years ago, that operation started crashing the filers because when some sensitive files were removed in preparation for replacement, the filer would crash and reboot even though it was a non-root volume that was being changed. I stopped rsync'ing and have not restarted that practice. We currently back up the /etc directories on all our filers to other systems. There may be someone on the list that can testify as to whether NetApp has made the non-root volumes more forgiving when it comes to /etc files being changed and removed. I didn't follow it for lack of time. At 1:59 PM -0400 9/23/03, devnull(a)adc.idt.com wrote: > > Some may say "Whoa, you have to have a separate root volume!" >> I say "Baloney!. >> The only time I came close to losing a root volume was in the infancy >> of NetApp; we did have to reboot and use a different volume as root volume >> (I had copied the /etc to the non-root volume) >Did you update your /RootVolume/etc/ to /NoneRootVolume/Another_etc_dir >daily using rsync ? Or was it a one time copy thingie ? > >I am quite sure i want to do this on my filers, there has been some >discussion on this list suggesting that it might not work. It has for you >and so any pointers are MUCH appareciated. > >> Another concern might be that "We're afraid the root volume will get full!" >> which you can easily avoid by judicious and simple quotas. >Agreed. > >> The only other thing I can think of is the time it takes to back up and >> restore by volume size. >Very true. > >If backup is done via nfs rather than NDMP, it might not be an issue >though. > > >Thanks -- }}}===============>> LLNL James E. Harm (Jim); jharm(a)llnl.gov System Administrator, ICCD Clusters (925) 422-4018 Page: 423-7705x57152

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

toasters September 2003