Clarification: I had bad experiences with disk reliability, I told different friends, that in turn had very bad experiences including multiple disks in RAID breaking in the same time range.
http://forums.seagate.com/t5/Barracuda-XT-Barracuda-Barracud...
So beware of having those installed, under high I/O they will fail at a random instant.
However, support is acting incredibly quickly, within 5 minutes of sending the support request they got to work on the server.
Because of missing data we asked for a data cable replacement first (instead of just replacing the disk) - and it worked. Even more, it works without problems since. YMMV of course.
BTW, I have worked with many providers but Hetzner is my favourite. Great support and very few problems.
> BTW, I have worked with many providers but Hetzner is my favourite. Great support and very few problems.
I'll second this. Support is/was often a horror story for me, but so far Hetzner has been absolutely impeccable.
Contrast this with Mediatemple, where I once had to engage in a lengthy support battle with a chain of incredibly ignorant and condescending support staff before being escalated to a senior person who finally, finally admitted I had found huge problem with the cluster. This took several days of intermittent outages, dozens of emails, and a lot of energy...
After the disk failures, however, some of the virtual images I had on there had trouble booting. They were complaining about bad blocks. After running fsck they ran again.
The whole experience made me realize that I really need to run on a filesystem with checksumming. Just plain RAID-1 only helps for a limited subset of errors, and only if it is able to detect which disk failed correctly.
Needless to say, rebuilding 3 TB raid took a long time and downtime and pain... but all in all, it was worth it, considering the prices.
I can also confirm support was quite fast, even if some of the replacement drives were not new and some even came with bad sectors lol ... such is life.
The DB server has pumped over 10 TiB of data in the last 80 days (mostly non cached) and the index is rebuilt constantly and well over 1.5 TiB in size.
Anecdotal I know but another data point to consider.
These days I just expect it to go bad, so I use ZFS within FreeBSD to watch for errors, and when they happen I ask them to replace a single drive, and then the other.
So far it has worked very well for me with no data loss. I've been fairly lucky in that both drives have't failed at the same time.
Also the procedure to get the disk replaced is insane and involves you dealing with disk IDs... It's a RAID, then should just check what is the broken disk and replace it, end of the story.
Of course, that doesn't make parent's claim any more true; just pointing out that this is probably not the reason.
Also be aware that their abuse department shoots first and asks later. I know of servers that were off the grid for multiple days without access because of abuse on a friday - everyone relevant went on weekend.
They _are_ cheap and that comes with a few trade-offs. Doesn't make the product bad, necessarily. If you are cash-strapped and need raw power, buy without hesitation.
Blaming providers for the inability to secure your server is incredibly irresponsible. Unless service was interrupted because Hetzner misread an abuse report for another server (which is highly improbable and most likely is not what you implied in your original statement), you shouldn't ever be pissed that your host took action to prevent abuse from occurring.
things that warrant disruption of service and are defined as abuse:
1) any type of DoS attack
2) phishing
3) malicious activity (ssh brute force, port scanning without consent, etc)
4) cp
5) anything that generates a court order with jurisdiction, including DMCA if you have anything to do with the US
6) spam/spamvertised site
there are more things that can be defined as abuse, but the others are more or less discretionary, imo.
The problem is that relevant persons assisting you in resolving the problem (e.g. giving access using a secure connection) are not reachable over the weekend and take their time. That is a problem of the provider.
If you ignore the 24 hour limit, it will just escalate the case to a support person, who then decides what to do. There is no automatic shutdown, especially not for multiple days.
Not being available for resolution is a different thing.