New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Man.....
everytime my HK vps on HH suddenly went offline, i worried about "data loss"
the same
Rclone to backblaze everyday and you won't worry..
that's why you backup
I had 2 TB server, all data lost. I was very happy with service, but now i am scared.
backing it up in backblaze will cost me more than double the price of VPS it self
Backup to Google Drive then
$7-$20/month unlimited storage
can't get past that daily upload limit 😬
750GB per day per account
Just make multiple (free) accounts, add them to a shared drive, and run sync operations via Rclone or just do incremental backups. Anyways, if you have a 2TB VPS, 750GB/day should be enough unless your data is constantly changing
still more than server price, 90$ 2 years blackfriday. Anyway I had backup of scripts and configs so now slowly restoring.
I run Minio, the egress is without accounting for deduplication.
I have 20TB of non replicated data, stored in chunks across each VPS. I can withstand one failure and be able to rebuild.
Just create service accounts. Easy work around, and if you need auto swapping there is a forked rclone that does it.
We don’t do Storage VMs in Hong Kong.
Our NVMe nodes have software RAID-10, and none of them use hardware RAID cards, which were the culprit in the last few incidents.
The rebuild times, when NVMe drives fail are usually in hours and not days due to the size and speed of NVMes.
So no - none of the issues we’ve had in the recent past would apply to them. That said though, please always keep backups of your data.
And on Storage nodes, we have mitigated this issue (or lowered the risk by a huge factor) by using multiple RAID cards and multiple RAID arrays per node, instead of single large arrays.
Not sure if your comment was meant more as a joke, but I hope the info above helps calm your nerves.
Hi,
mine is Promotional Package NVMe 4G (E5)
Thanks for clarifying, i don't mean to jokes , sometimes uptimerobot unable to ping my HK server (down and then up again), i guess bcs the network unreachable or rto for a while.
But when i logged in to ssh, uptime still on the track (75days+). So, yeah, my vps run smoothly.
LowEndStaticians, is this true?
Adding more hardware RAID cards increases the failure rate and reduces MTBF. It sounds like decreasing the array size is intended to speed up rebuild times, but is that actually true? Is the rebuild time for a 10TB drive the same in a 4 drive vs 8 drive RAID 10 array? It's still writing 10TB in both cases.
Or is this halving the data loss rate by reducing lost data by 50% when failures occur?
I thought risk usually goes down as disk count is increased (probability or something spread over more units?)? I never took statistics in high school and regret that.
Smaller arrays do rebuild faster, yes. Smaller redundant arrays also give them a better chance of recovering from inevitable drive failures. Risk goes up as disk count is increased because the likelihood of one drive failing is always the same. This is the reason why RAID 0 actually gives you negative redundancy, since the chance of one drive failing is now multiplied by however many drives are in the array. Multiple small arrays limit your exposure to one drive failing though, instead of rebuilding a 30 drive array you might just have to rebuild a 5 drive array. Again, HH is acting on pretty good faith here, no need to jump down their throats
delete
very unrealiable lately because of one failure?
because one node stop working? isn't that normal?
One? What are you even talking about?
In very short time 3 different locations failed. Not to mention numerous complains all over the forum about unanswered tickets for weeks.
As much as I liked them in past yes, they became very unreliable.
Remember we had a provider for which people kept finding excuses (including me) all because of the low prices and offers? This is how the term "involucrated" got created around here.
one failure? they can't fix the network in Chicago for many months. It takes them at least six months to go from ignoring to realizing the problem
this is how it looks right now
Correct me if I'm wrong but I don't think the word "involucrated" "got created" around here.
It was just a word cociu has been using for several years in his replies/communications
True. And we adopted that term after he closed business and services without explanations.
To be fair, 3 different nodes in 3 locations failed. Considering how many they run, it seems somewhat expected, although not good. I run several dozen storage servers personally and I lose around one node every few years. Drive/hardware failures happen. About the tickets, I'm sure it happens, but every ticket I've opened is answered same/next day. I don't understand why people are having this issue, I've found HH support decently fast. Not denying there are issues, I've just experienced the opposite.
As far as I've seen, the trick with HH is to 1) have an actual issue, 2) collect all tech info that you can, and 3) wait for 16-20 hours and then file a high priority ticket directly stating the impact (and no, it can't be "losing millions").
I once had a routing issue which was promptly solved this way, though HH never replied to my ticket.
How many of those "tricks" worked for you to be opinionated about the case? Just curious.
I just try to be explicit in what needs to be fixed eg: "Server IOwait high on instance IP address" and not just "server slow fix now". Including as much data as possible into your ticket just seems like good manners at this point, anything less feels like an insult to your provider's time, especially if you're on a promo plan. The last time I had an issue I got a ticket response in under 5 minutes with some advice from Mike (love that guy). Expecting your provider to do detective work on your service that doesn't give them any margin is insulting and it's no wonder why those types of tickets take longer.
How do you know that all those people (and there's really a lot of them complaining recently) expected from a provider to do detective work for them?
Just because of you personally have good experience SO FAR (you may change your opinion anytime) this don't mean that other ticket requests are less legit or helpful in description of a problem than yours, so please just stop making up things about other people support tickets ...
Just work a tech support job at any company. The number of customers who refuse to cooperate or provide any information and yet demand that their issue be fixed "RIGHT NAO" is the primary driver of attrition in those jobs.
Or maybe, you could create a new thread with the contents your tickets for us to decide where you lie. I'm expecting a bellicose answer to this request though
Have done work with webhosting providers for the past 6 years. Customers are just like this.
And somehow you can ignore good experiences with hosthatch support simply because some people have been complaining? The reality is the most vocal people tend to be the ones who want to complain about something.