djbdns + svn + cron = dns replication

I'm sure this has been done before - probably even pretty often - but since I couldn't find anything on how to make it work properly without spending too much time on it, I decided to write up what I did there.

I wanted to replicate my dns-database between several djbdns installations on different operating systems. When I first set this up, I went for ssh without password. there were some minor issues with that:

- no passphrase (must be something spiritual but I don't like that)
- without rsync involved: one-way only (as in master->slave, slave...)
- with rsync involved still not optimal to keep env, root etc. apart
- no versioning, no roll-back, no safety

on the plus side it's been super-easy to set it up, though.

since I recently bought into a new server plan and just a month before found and old wrap-pc that had been configured as openbsd-dns-server years ago, that was still working, I decided to put some more brain into the matter. these were the results I was looking for:

- it shouldn't matter where to change configuration
- shouldn't be needed to log into the dns server ever again
- versioning/backup/roll-back should be an option
- only parts of the directory should be replicated/distributed
- take into consideration that not every server is the same
- server should pull the updates
- encrypted communication
- authentication required for at least writing

the main trigger was going to be cron like before but I had to find something for allmy other needs. subversion was the solution.

I finally went for a subversion repository with different env-directories for each server as there are IP-address as well as the path to the root-directory configured and those can differ from server to server.

the svn is behind an apache2 just because it's already been there and I couldn't be bothered to make svnserver fly with ssl and authentication. apache works just fine. so we have svn behind apache/ssl and that part took my rusty brain two attempts and maybe an hour to get it working properly. I'm getting too old for this stuff.

the configuration of the servers didn't take much longer and finally writing a script for cron with all the necessary environment variables took another two hours. I bet every half-way routined administrator finishes this in less than an hour.

now let's talk source (and yes, I'm perfectly aware of there being room for improvement - so just let me know if you find something, will you?)

since djbdns/tinydns has become part of the even debian lenny distribution I think you will find the basic setup instructions for your favorite flavor of linux/unit real quick on any major search-engine. so let's dive into the specifics of my setup, will we?

first of all I'm kind of a control-freak when it comes to my servers. I didn't really like to have all domains on one data-file and decided to add a directory for configurations I'd later just "cat > data". also I prepared some files I was going to work with later.

cd /service/tinydns/root/
mkdir data.d
touch tosses.info.data
touch lsrv.net.data
touch make.sh
touch svnup.sh

I went on configuring my .data files like I'd do it for a regular tinydns data file. to assemble the data file into the root directory, I just hacked a little make.sh script like this:

vim make.sh

#!/bin/sh
ROOT=/service/tinydns/root
DATAD=$ROOT/data.d
echo # This file will be generated by $DATAD/make.sh > $ROOT/data
cat $DATAD/*.data >> $ROOT/data

chmod +x make.sh
svn add make.sh
svn ci -m "build your data from data.d with make.sh"

my dnsservers were half-way done. next up had to be the subversion configuration at the svn-server as well as at the dns-servers.

at the subversion server I made a repository just for the dns servers. into that repository I made some directories and finally I added some basic auth ssl configuration to the apache setup:

mkdir -p /var/svn/repositories
chown www:www /var/svn/repositories
su www -
svn create /var/svn/repositories/dns
svn mkdir --parents /var/svn/repositories/dns/tinydns/root/data.d
svn mkdir --parents /var/svn/repositories/dns/tinydns/ns1/env
svn mkdir --parents /var/svn/repositories/dns/tinydns/ns2/env
svn mkdir --parents /var/svn/repositories/dns/tinydns/ns3/env

now I had the directories for data, data.cdb and the .data files as well as env directories for each of my dns servers.

the apache2 configuration was pretty simple as well. after enabling ssl and configuring the ports as well as the virtual hosts, I added the site configurations for my dns repository as follows as root:

vim /etc/apache2/sites_available/svn.tosses.info

<VirtualHost 192.168.12.34:443>
  ServerName svn.tosses.info
  ServerAdmin svn@tosses.info
  SSLEngine on
  SSLCertificateFile /etc/apache2/ssl.crt
  SSLCertificateKeyFile /etc/apache2/ssl.key
  Alias /svn/ /var/svn/repositories/
  <Location /svn>
    DAV svn
    SVNParentPath /var/svn/repositories
  </Location>
  <Directory /var/svn/repositories/dns>
    AuthType Basic
    AuthName DNS
    AuthUserFile /etc/apache2/htpasswd.svn
    require valid-user
    order deny,allow
    deny from all
    satisfy any
  </Directory>
  ErrorLog /var/log/svn.tosses.info-error.log
  CustomLog /var/log/svn.tosses.info-access.log common
</VirtualHost>

htpasswd -c /etc/apache2/htpasswd.svn dns

after a graceful restart I could access the repository and see the directories I had made before. my work on the svn-server was done.

back on the (till then) main dns server I checked out the empty directories and started adding files to the svn:

cd /service/tinydns/env
svn co https://svn.tosses.info:443/svn/dns/ns1/env .
svn add IP ROOT
svn ci -m "IP and ROOT configuration for ns1 added"
cd /service/tinydns/root
svn co https://svn.tosses.info:443/svn/dns/root .
svn add data.d data.cdb
svn ci -m "general data.d directory and data.cdb database for all dns"

so far so good. for cron I still needed an svn update script including the necessary variables. those scripts needed to be executable and I had to add svnup.sh to the crontab:

cd /service/tinydns/root/data.d
vim svnup.sh

#!/bin/sh
PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
LOGNAME=root
USER=root
HOME=/root
cd /service/tinydns/root
svn update

chmod +x svnup.sh
svn add svnup.sh
svn ci -m "now svnup.sh will update the svn"

crontab -e

* * * * * /service/tinydns/data.d/svnup.sh > /dev/null

for the other two servers all I had to do was to repeat the checkout of their own env directories and in the root directory I needed to remove data, data.d and data.cdb before a checkput could be successful. but those had to go anyway:

on ns2

cd /service/tinydns/env
svn co https://svn.tosses.info:443/svn/dns/ns2/env .
svn add IP ROOT
svn ci -m "IP and ROOT configuration for ns2 added"
cd /service/tinydns/root
rm -Rf data.d data data.cdb
svn co https://svn.tosses.info:443/svn/dns/root .

crontab -e

* * * * * /service/tinydns/data.d/svnup.sh > /dev/null

on ns3

cd /service/tinydns/env
svn co https://svn.tosses.info:443/svn/dns/ns3/env .
svn add IP ROOT
svn ci -m "IP and ROOT configuration for ns3 added"
cd /service/tinydns/root
rm -Rf data.d data data.cdb
svn co https://svn.tosses.info:443/svn/dns/root .

crontab -e

* * * * * /service/tinydns/data.d/svnup.sh > /dev/null

now every minute every server tries to update from the repository. sure, that's a lot and maybe I even go back to five minutes as checking interval.

the only thing that might become an issue is if I change files in data.d on several servers without immediate check-in. but there are enough ways to fix this in svn.

Faulbaer (I hope I haven't forgotten anything)tt

if in doubt ... use xen

it's done, I've decided and it's become a xen system on lenny. to sum it up real quick: ubuntu server: pain in the arse. kvm very much unfinished busines. vmbuilder hyped crap. virsh pretty much is like masochism.

if I hadn't spent all my energy on trying to get virsh and kvm running on ubuntu first and later on debian lenny - well - I might have given centos a try. but after two-and-a-half day I was just too exhausted for further experimenting and just worked with what I knew would work: debian lenny, xen and lvm. xm-create-image works like a charm and does what it promises. not half of it or only two thirds right and the rest is a mess - it just does what it can and that it does very well.

in fact I already started migrating services to the new box. the weblog is already there. I couldn't be happier. the really big thing is going to be the zimbra migration but I've done that before and it worked every time. so no bad expectations here.

I've decided to drop ejabberd since I'm the only user it's kind of an overkill. also I guess zimbra's jabber services should have improved by now - so I'm gonna try that.

I haven't yet decided if I'm going to upgrade the old server or use the old machine as backup system. will figure that out sooner or later - got enough time for thinking, now that the new system is running well and migration seems to be going on pretty easy, too.

firt blogpost on the new machine - hope it works - fingers crossed ;)

Faulbaer (permissions set properly? we'll see ...)

kvm? I've had it.

I've really had it - I'm going back to xen - kvm may be the better, faster, more stable and more actively developed engine but to be honest, it's not just the engine, I need the tools around it to be working, too.

I've tried to make it work for one-and-a-half days now and that's kind of the longest time I've put into new software-product that is constantly failing, ever. kvm may be working like a charm but vmbuilder and virsh are nothing short of a pile of crap, designed to make a seasoned administrator cry.

vmbuilder cannot resolve dependencies and appears to be so much more beta than the straight-forward xen-tools I cannot begin to understand why I put so much effort into trying to make it work. I should have just stopped when it wouldn't build a 'hardy' guest. I should have stopped, when I saw the --part param would only accept four partitions - all of them turned out primary by the way. I should have stopped when I saw that --raw was just constantly ignored. I should have stopped, when it automatically added the wrong bridge device. But I tried and tried and tried - just to give virsh a go.

virsh is so overdeveloped bloat it makes me want to organise it's developers into an anonymous xml-java-addicts group. not only is it impossible to see where it's current configuration is stored - some parameters cannot even be changed without going into the virsh shell first. this poorly documented pile of shit wouldn't see the changes I made in qemu/networks/default.xml even after a reboot. that's why most tutorials work with the default settings for virbr0 - their authors just didn't know where to change this shit. Ididn't find even a single document describing where this had to be done - not one! attaching a physical drive always ended up in crashing the guest to never be started again because virsh actually messes up it's own xml config! so you need to edit the machine to make it work again. then it starts but hey - for how long?

and don't get me started on the idiotic qemu drive images! could someone explain to me why the xml file sais the image was going to be mounted as hba/ide but really it is going to be mounted as scsi sda? what sense does that make? is there a way to handle this properly via lvm? no, there isn't. because all of your lvm volumes will end up to be represented as drives, not as partitions which would make plenty more sense to me.

I'm through with this shit - I'm going back to lenny with xen - I cannot even say it's been nice while it lasted because it hasn't.

Faulbaer (fsck that shit - I'm outta here!)

blogging with blosedit

finally got around to installing some slightly more appealing frontend to my blosxom weblog than bxr. mainly because there doesn't seem to exist any good blog editor like ecto that works well with the metablog engine and doesn't cost a fortune. I guess that's because there aren't many engines besides wordpress and serendipity out there anymore ... ah, yes movable type, almost forgot about what this blog had started on.

at last I can blog from everywhere with wifi and soon it's going to be wifi or any given cellphone standard.

Faulbaer (so exciting)

ipad, iphone and the pre

I'm really looking forward to the ipad. not only that but actually signing up to become an iphone/ipad developer again - although it didn't get me too far last time I believe this time it's going to be different. there are things missing from the ipad - things I need ^_^

the only downside there is to longing for the ipad right now would be to be european as there neither is an official release-date for even the entry-level ipad nor is there a price. also there is no official word about the 3g version not to be sim- or netlocked - so let's hope for the best.

for some days or so I considered to buy the basic ipad although I'd rather be completely mobile and independent from tethering - not that my uk-based jailbroken blacksnowed iphone would tether properly anyway - for that I can only use my palm pre. no - it's going to be the 3g version. I have a 'spare' o2 data-flatrate sim that I got with my little asus eee pc netbook that the ipad is going to replace.

a propos palm pre - apparently they're broke - again. this is a major disappointment but not so much of a surprise to be honest. it's not that the pre didn't have potential or that the marketing was all wrong. even (at least in germany) they picked the perfect partner for the pre. but there were some major issues that added up had to lead into disaster:

- the device had been oversold, the pre never really came close to the iphone in terms of hardware or software and it couldn't really do what had been advertised

- promises were broken when the pre didn't connect to itunes - it's not as if there wasn't a way to get information out of itunes, the palm people just didn't get it done. why the hell didn't they supply their mac-users with a proper sync-app for the pre?

- communication issues and jobs only partially done - they shipped far too early but didn't really fix the device afterwards. I still can't properly backup my pre to my mac or to the network or to anywhere. the touch-stone charger is a great idea but without wireless synchronization it becomes an annoyance. even bluetooth protocals only have been implemented half-way. I still can't browse my pre's filesystem via bluetooth - why not?

- quality has been a big issue with the device as my battery is almost always dead after less than a day and the device has been replaced two times after dying on me - within the first month as I recall. also the device doesn't feel properly built, the keyboard-layout could be much better and the responsiveness as well as the precision of the touch-screen are less than mediocre

- the software did introduce some good ideas but since most of those great features hadn't been implemented properly the device couldn't be a success. both - iphone os as well as android have caught on with web-os. there isn't much left only the palm can do - and whatever challenges the competitors solved they did a better job at it than palm did.

actually I think the whole approach might have been flawed. it's easy to conclude that afterwards, I know that, but I think what palm should have done to compete against the iphone as well as against android would have been a much more open policy. they should have aimed applications at communication like sip-telephony, skype, jabber and maybe even some clients for the major blog-platforms. also they should have marketed tethering earlier and the quality should have been on par with at least their past products like the palm Vx or the palm tungsten or even better the ones of their competitors, the iphone or the ipod touch.

they should have opened their store right from the beginning and there shouldn't have been delays in shipping os-updates to outside the united states. why palm closed down their us-store to european customers is also an unsolved riddle to me. making developers pay for submitting applications to the store still doesn't seem right to me.

I'm very interested into who is going to buy the pre and what the next owner is going to do with web-os, the pre and how they are going to try to compete with apple and google. traditionally palm is a closed platform like apple has been. their attempts of licensing the palm platform didn't work out too well. palm might have missed their window of opportunity to become a strong competitor to apple on their locked-in turf but they still might have a shot at the open market where google is king of the hill right now - in the open market it's ok to just be good enough for a lower price as linux and windows have shown us over and over again - there nobody needs to be perfect.

Faulbaer (cheap but acceptable is the new perfect)

os-jihad: which flavor of linux for my next server?

my server is getting old - not only the hardware is way beyond due but also the operating system needs to be at least upgraded to the next major version. as always in this situation I'm pondering my operating system options - what has happened since my last major upgrade? what are my current options and what flavor of unix or linux do I want to use - and why?

I have to consider several requirements for my operating-system shootout and I have to decide between at least those competitors:

- debian lenny
- ubuntu server
- centos
- gentoo

my server currently is running xen, apache, varnish, djbdns, zimbra, ejabberd, mysql and occasionally some ftp-daemon. I could switch to lighty, I could switch to bind I even could switch to kmu or some other alternative to xen. but usually changing less works better for me.

right now I'm torn between ubuntu server which is supported by the zimbra folks and gentoo which I really like to work with because it's cutting edge and the package maintainers really keep their shit together most of the time. debian lenny would be my first choice if it worked with zimbra right out of the box and without any hackery involved. centos is just not my flavor of linux but still a sane choice in regard of stability, long-time-support etc. ... well - I'm not always a sane person I guess.

since I know so many great sytem administrators with their own perfectly good reasons for favoring one flavor of linux over the other I'm convinced I'd better just pass the ball to you my dear friends and let you recommend your very favorite linux to me.

please try and keep the signal/noise ratio in check and just suggest operating systems that really make sense to me. I can't afford vmware infrastructure nor do I own hardware that is certified to be used for vmware esx. if you really feel the urge to correct any commentor or me for that matter, please do this in a civilized manner.

at least state the three big "w":

- what? (which os would you suggest?)
- why? (why whould you suggest this?)
- when? (is a major change expected that's worth waiting for?)

please help me find the right operating system for my next server and leave a comment.

Faulbaer (guess I'm better gonna twitter this, too)

gulp stinkt mir

seit anfang des jahres bin ich bei gulp dabei und waehrend ich es schon fuer eine frechheit halte, nach einer sehr kurzen und nutzlosen testzeit, in ein abo gezwungen zu werden, bin ich als kunde des auftragsportals auch noch in anderer hinsicht unzufrieden. gulp stinkt mir - bis hier hin!

die seite kostet jene, die projekte suchen geld, ebenso wie solche, die projekte einstellen. ob 25 euro monatlich angemessen sind oder nicht, darueber laesst sich sicher streiten - ich finde es gemessen an der geringen menge eingehender projektvorschlaege zu teuer - auch liegen die vorgeschlagenen projekte zumeist so weit daneben, dass ich ernste zweifel an den faehigkeiten der entwicklern hinter den matching-algoritmen habe.

hinzu kommt, dass ich meine firma dort nicht angemessen repraesentieren darf, denn alles, was auf meine identitaet schliessen laesst, ist mir per vertrag verboten bekanntzugeben. damit soll ausgeschlossen werden, dass mich etwaige auftraggeber direkt ansprechen, ohne gulp dafuer ein angemessenes vermittlungsentgelt zu leisten. als vermittler kann man gulp natuerlich auch sofort vergessen, weil gulp ja selbst als vermittler auftritt und somit im pitch immer die nase vorn hat - dadurch suchen natuerlich vermittler auch nicht auf der plattform nach mir, dem experten.

was mich allerdings jetzt schlussendlich mehrfach genuegend veraergert hat, als dass ich mich zu einem blogeintrag genoetigt sah ist, dass mir gulp nicht nur unpassende angebote unterbreitet, sondern darauf, unter androhung der einseitigen kuendigung meiner mitgliedschaft, mir auch noch eine reaktion abnoetigt - eine reaktion, die dann postwendend mit einer projektabsage beantwortet wird. das ist gleichermassen frustrierend, wie unnoetig und ich habe ehrlich gesagt kein verstaendnis dafuer, wieso ich als kunde/mitglied/nutzer des portals mich so behandeln lassen muss.

ich zahle fuer sechs monate eine frech hohe gebuehr von 150 euro, werde dafuer schlecht vertreten, schlecht behandelt und erhalte unpassende angebote, auf die ich reagieren muss, was meine zeit kostet, um nicht von der plattform zu fliegen. davon abgesehen befindet sich das gesamte portal technisch noch in der internet-steinzeit - das kann ich mit meinem hintergrund im design und betrieb von modernen portalen wie neu.de und pkw.de ganz gut beurteilen.

fuer mich fuehlt sich das nach abzocke an. gulp hat jetzt noch knapp drei monate zeit, mir passende projekte anzubieten, dann bin ich da weg. weil es mir da so stinkt, kuendige ich aber vorsorglich gleich heute zum naechstmoeglichen zeitpunkt - kann die kuendigung ja aufheben, falls sich noch etwas aendert - wovon ich wirklich nicht ausgehe.

Faulbaer (hat hier irgendwer gute erfahrungen mit gulp gemacht - wer den ich kenne?)

palm pre app store - don't pay for nothing?!

received a mail from palm today inviting me to download some electronic arts titles for free. after palm had removed all the us titles from the german store, I hadn't spent much time in their 'app catalog'. it's full of apps again and some of them seem to be worth a look.

what strikes me as odd idea is that still everything there is for free. either palm doesn't yet have a payment-model for their platform or it's just not yet switched on. I don't get that. from my point of view this is going to annoy customers in the long run because as soon as they switch on payment, the old customers won't buy apps because they already have what they need or just don't see a point in paying for things they used to get for free. the new customers won't pay either because why should they pay for apps other got for free? when we look back into the history of radio and tv, that's what happened there - and that's what made it so very complicated for pay-tv to get their business going - at least where I live.

anyway - I'm downloading tons of apps right now, so the 'freebie' newsletter worked to get me into the app catalog again. we will see if this is going to be enough.

while I'm at it, I must admit that web os has gotten better with the latest update (1.4.0). also some of the shipped apps have improved a bit so the palm still is a valid alternative to the iphone as well as android phones. I like the new camcorder option in the camera app and all those little tweaks, fixes and features palm added to the phones's current software.

syncing over the air is still missing - I don't understand why that is. it would be easy to add some rsync-based service via samba or sftp - why doesn't palm add such a feature? it's nice to charge the palm wirelessly so there should be an option to sync it the same way, too.

the only thing I don't expect them to fix via software upgrade is the messy touch display. it just cannot compete against apple's device. but from what I know - nobody else offers a display coming even close to what apple sells.

what still doesn't seem to work is the task synchronisation between zimbra and the pre. tasks I add on the pre find their way to the zimbra desktop but what I add to the zimbra desktop sadly doesn't show up on the pre - interestingly apple's mail.app won't show zimbra's task but the one of the pre. but tasks I add in mail.app won't show up in either the pre or the zimbra desktop. besides that it's taking ages for all the tasks to sync - this is just not ready for prime-time.

what really anoys me most on the pre is that it's localization is all wrong. when I set my pre to 'english' I expect it to speak english and get all the apps in english language - but it doesn't. at least the electronic arts apps are partly in german, partly in english - this is really poor performance. oh - a propos performance - it doesn't quite make a good impression if your game stops playing music or doesn't continue loading the menues after a game of need for speed - since I dind't have to pay for the app I don't mind too much but as a paying customer I'd expect flawless game-design customized to the very abilities of the palm pre - at least my pre isn't strong enough for need for speed - not even in single-tasking mode.

I'm still considering to sell the pre but since it sells for only €250 used I guess I can just keep it and watch it to further develop into the iphone killer it set out to become ;)

Faulbaer (not really.)

model communities, forums and portals

I'm not impressed. not only that I expected far more photographers on an international portal like model mayhem, the defacto leader of the model photography portal pack, for me as a newcomer next to everything in these web-services feels wrong and yesterdayish. almost all of the user-interfaces appear to be hurried together. most of them look like they had begun as a web-shop, a forum or even a weblog-engine. there is no interconnection, no way out and no way in. there is no way to standardize your input or to update all of those portals at once. there seem to be no APIs anyone can use. there are no iphone apps to work with those sites. the only way to stay current is to keep a tab open in your web-browser.

since all of them claim they'd make our lives as photographers or models or stylists easier, I see some great opportunities ignored here. portals that look like shit, feel like shit and don't have any means to improve won't be able to survive too long.

I guess there will be one to rule them all in the near future. It's key-features will be the following:

- connections to all the important social networks like facebook, flickr, google, twitter, soup.io and so on

- a simple user-interface, streamlined to the needs of the users. models get a model view, photographers get a photographer view etc.

- import-wizards to get your information (and hopefully your contacts) from other sites

- it will be invitation-only for a time

- it will offer spam-detection as well as counter-measures

- it will offer add-words or some similar revenue-stream again for each user individually

- it will offer a recommendation-system for users who interact together in the way facebooks suggests contacts and amazon suggests stuff based on previous interactions

- a subscription will probably make advertisement go away

- with any luck there will be localized standard-contracts for tfp or other shooting-types ready to be agreed on online, printed out and filed before the shoot

- the service will be meshed with everything. images will be hosted at flickr, contracts in google-docs, mail with google, comments with disqus, locations with qype and google-maps, stuff with amazon, ebay and google, contacts with facebook and twitter, paypal for payments etc. why invent the wheel over and over again? most problems have been solved an just need to be pieced together via api-calls

- there will be an iphone-app with notifications to never miss an opportunity

- there will be filters for people with many contacts to keep their inboxes in check

- maybe there will be forums attached to the service but those need to be moderated and that seems to be overkill

I'm not yet sure where to host nude-photography properly since flickr isn't really up to the task for various reasons. maybe those images would be the only ones hosted on the portal.

anyway - I'm pretty sure there will be a new big player in the market within a year and I'll be happy to switch to their service if they even got half of what I proposed earlier.

Faulbaer (currently registered with 5 such services all of which suck plenty)

why I cannot reccomend to buy the drobo anymore

I've been having several problems with my drobo on several occasions. it's not that it's a bad idea or a bad design in general - it's just that in most cases for me it doesn't do the job as advertised and it doesn't help me with the problem that made me buy a drobo in the first place.

let's get into the details, shall we?

let's start at the beginning, while we're at it ^_^

in the beginning there were files and they got copied and changed and copied and change and their numbers and sizes grew and then space just ended and there was no more room for all those directories full of files ever growing in numbers and sizes - and the solution was to get more space, another hard-drive ... and as you might have expected the new one became too small soon again and the next and the one after that and then the first drive failed and I lost a gigabyte or two and I started backing up and keeping versioned redundant repositories on encrypted disks and/or disk-images and suddenly I needed twice the space and in some cases even more ... and I didn't know how to organize all those drives and the repositories and I actually considered deleting everything and starting all over again - not for long because that was around the time drobo launched a neat marketing campaign targeted and perfectly customized to my very needs - it promised to be the solution to all my problems and it did it with a cute girl explaining everything in a well made video on the internet where I live.

I'm not new to the business and I didn't believe everything I saw in the video right away. I did some research and read all the information data robotics provided me with. I reckoned they were cheating a little bit here and there and I expected that the important information was going to be the information they had left out in the video which by then had gone viral - not completely without my help and the help of so many others desperately searching for a solution to their storage problems.

I have a reasonable amount of experience with storage in almost every shape there is. I setup and maintained small, medium sized and large raids - even small and medium sized sans although by the time I decided to buy the drobo it had just been large raids and nas-storages.

all the information I got and all the pondering couldn't spoil my expectations of the drobo which were very high to say the least. it looked as if it was going to be the real deal and although I wasn't going for the 1.0 usb version I totally went for the firewire 800 drobo which proved to become a major disappointment after all.

broken promises

the video had shown a movie continually playing while expanding a drive or recovering a volume to a replaced drive - that is actually not a lie but they could have mentioned that a volume of the size of 4 tb was going to be recovered over days, not hours. if you wanted it to finish recovery within a week you better not touch the drobo while recovering.

the video also hadn't shown that there was actually a size limitation to a drobo - you had to decide on it's volumes' virtual size in the setup process. to date there is no way to grow the drobo beyond that. the maximum size is 16 tb in my firmware.

also there is no way of uploading a new firmware to a broken hard-drive in the drobo which is really a shame. if you are as unlucky as I was you end up with removing and upgrading your hard-drives one-by-one, always recovering and rebuilding the drobo volume.

the build-quality of the drobo is hmkay - it's loud and plasticky. I lost the cover-flap of one drive-bay by removing said drive for the mentioned firmware-upgrade. it looks rather ugly after some weeks of collecting dust. remember: the drobo wasn't meant to be used in a virtually dust-free data-center but at home or in an office where there is dust - tons of it!

I also don't like the way the sockets have been placed or that it has a separate power-supply instead of a proper 115/230 volt connector.

what drives me nuts all the time is that I can't tell the drobo to NOT shut down the drives, to not send them to sleep while not being used. the caching on the drobo seems to be shitty, too and whenever my mac wants to access data on the drobo I hear the spin-up of the drives inside taking ages and after a rather long while I can use the mac again. granted, half of that issue is apple's fault but hey - if those drives didn't sleep the other half wouldn't be such an hassle.

the overall performance of my fw-800 drobo is just poor. I haven't had any equally slow firewire drive after firewire was born. I don't actually get why that's the case. in theory there should always be at least two places to read data from, making the drobo at least double in read-speed of the slowest drive installed - but the reality is that it more like seems to be half the write-speed of the fastest drive installed for reading and a third of the write-speed of the slowest drive installed for writing data.

also accessing data in parallel proves to be problematic if not dangerous. I'm used to move data around quite a lot. I copy this and that to the drobo while reading some and more from it. the drobo sometimes responds so badly to that kind of usage that I tend to reduce all of them to just one job of either reading or writing - that is a big disappointment since I have four drives in that drobo and working with four terabytes of data involves parallel reading and writing a lot.

besides broken promises the drobo itself fails from time to time. that's when you send a mail to their very nice and very well trained support-team. and sure enough they could help me most of the time and I didn't actually lose any data but the drobo failing isn't really an option and taking recovery-times and bad performance into consideration it is kind of a deal-breaker for me, too.

I'm looking into alternatives to the drobo and deep inside me there is a feeling building up to just get rid of the drobo while there is time left - with the next major crash I'm going to switch to a proper networked storage or maybe even back to a file-server. the drobo let me down too often at too many occasions, didn't keep it's promises and can't help with the main issue: logical data-corruption.

the drobo can only fix broken drives and that it does slowly and clumsily. for it's prize-tag this is not enough and I'm not going to buy more drobos for further versioning and redundancy. the device is just not good enough for that and I lost most of the trust I put in it almost a year ago.

so - no, I can't recommend to buy a drobo for my use-cases. it's not that it isn't ready for prime-time - it's just a really bad performer and the only thing it does well, it doesn't as soon as you outgrow it's limitations.

Faulbaer (any recommendations what to buy next?)