How to achieve Less downtimes ?

Chievo · May 17

I was thinking about to lower the downtimes somehow in a shared hosting plans. Is some provider offering a paid way to do it? I was thinking for DNS way or load balancer . Which is your advice? I prefer the shared hosting than running VPS for example.( Not about the cost but for delegate it ) .

Edit:
VPS buy different plans+ proxmox Ha cluster etc and rsync the content

It is another option as well

euronodes · May 17

Keep the copy with 2-3 providers and loadbalance via cloudflare? Thats self-oiling machine.
If you utilize the DB, PostgreSQL has the most stable replication out of everything we ever tested, incl Oracle
That really depends what you want to HA - different strategy for a blog, different for high-transaction CRM

"VPS buy different plans+ proxmox Ha cluster etc and rsync the content" - that s complete mix'n'mash that doesnt make much sense IRL.
How a wise man once said: sounds good, doesnt work
First of all is single provider - proxmox cluster will not work over WAN (technically it will but establishing cluster ring over WAN is pure nonsense - corosync requires 5ms max before everything turns into grey question marks)

Chievo · May 17

@euronodes said:
Keep the copy with 2-3 providers and loadbalance via cloudflare? Thats self-oiling machine.
If you utilize the DB, PostgreSQL has the most stable replication out of everything we ever tested, incl Oracle
That really depends what you want to HA - different strategy for a blog, different for high-transaction CRM

Thanks for the answer. Recently I have had downtime not critical let s say but if could be avoided or minimised would be better definitely.

euronodes · May 17

I edited the reply to address your direct question
Also theres now new phenomenon that few months ago Cloudflare initiated first time: DaaS

Thats Downtime as a Service. They failed few times over the month so badly and took down so much, that it was just widely accepted and everyone just shrugged and moved on.

"CF down? A OK, lets have a break, half of the world doent work anyways" - its almost global banking holiday.

We didnt use CF at that time and proudly announced that OUR INFRA STILL WORKS!
Which had as much sense as opening McDonalds on Mars. Works? Great, but what for? Nobody will come

So dont overkill yourself with that, get CF and be done. It doesnt make operational sense in outliving them - you will stay alone on this planet at the cost of oversized maintenance for the sake of maintenance

Chievo · May 17

@euronodes said:
Keep the copy with 2-3 providers and loadbalance via cloudflare? Thats self-oiling machine.
If you utilize the DB, PostgreSQL has the most stable replication out of everything we ever tested, incl Oracle
That really depends what you want to HA - different strategy for a blog, different for high-transaction CRM

"VPS buy different plans+ proxmox Ha cluster etc and rsync the content" - that s complete mix'n'mash that doesnt make much sense IRL.
How a wise man once said: sounds good, doesnt work
First of all is single provider - proxmox cluster will not work over WAN (technically it will but establishing cluster ring over WAN is pure nonsense - corosync requires 5ms max before everything turns into grey question marks)

So basically 3 VPS in the same data center you mean for the proxmox HA. ( So less than 5ms between them) ( Sorry I have not done proxmox yet ) It is a new for me . Thanks for your advice

euronodes · May 17

What do you want to HA? Some platform, website, eshop? Strategy depends on the product TBH

Chievo · May 17

@euronodes said:
What do you want to HA? Some platform, website, eshop? Strategy depends on the product TBH

Configuration is pretty simple woo+wordpress website. Transactions as well content as well.

euronodes · May 17

OK, so thats not simple, woo is not stateless content
Multiple provider will NOT do it without trickery like

active+passive replication (just failover without load balancing)
multimaster DB cluster (Galera)

IMHO you need to stay with beefy provider. Multi-provider HA for woo is more a project than a setting

And definitely I would avoid VPS and keep shared hosting for that setup. Everything else will tax your time

somik · May 17

@Chievo said:
I was thinking for DNS way or load balancer . Which is your advice?

Both. Use DNS to spread your "site" across multiple regions and implement a load balancer in each region. Load balancer should contain about 3 nodes per region. That is how most major companies implement high availability.

Also setup a monitor for all of them that notifies you if there are 2 missed pings from the nodes so you can take action before your client notices any downgraded performance. This way you can minimize your downtime to near zero.

Still things to note are that if the DNS fails, your entire stack will be useless (happened to cloudflare a few times, right?). If your load balancer fails, that regions nodes, even if online, cant respond to clients anymore. Finally, syncing data between all the regions is a pain in the rear...

Chievo · May 17

@euronodes said:
OK, so thats not simple, woo is not stateless content
Multiple provider will NOT do it without trickery like

active+passive replication (just failover without load balancing)

multimaster DB cluster (Galera)

IMHO you need to stay with beefy provider. Multi-provider HA for woo is more a project than a setting

And definitely I would avoid VPS and keep shared hosting for that setup. Everything else will tax your time

Mmmm it is difficult I guess proxmox cluster HA would be a "viable " option.

Chievo · May 17

@somik said:

@Chievo said:
I was thinking for DNS way or load balancer . Which is your advice?

Both. Use DNS to spread your "site" across multiple regions and implement a load balancer in each region. Load balancer should contain about 3 nodes per region. That is how most major companies implement high availability.

Also setup a monitor for all of them that notifies you if there are 2 missed pings from the nodes so you can take action before your client notices any downgraded performance. This way you can minimize your downtime to near zero.

Still things to note are that if the DNS fails, your entire stack will be useless (happened to cloudflare a few times, right?). If your load balancer fails, that regions nodes, even if online, cant respond to clients anymore. Finally, syncing data between all the regions is a pain in the rear...

Thanks for the advice . Which would be your advice for syncing ? On what technology I should focus on? Thanks

matserix · May 17

@Chievo said:

@euronodes said:
What do you want to HA? Some platform, website, eshop? Strategy depends on the product TBH

Configuration is pretty simple woo+wordpress website. Transactions as well content as well.

Tho not simple to setup high availability.

A 3 node Proxmox cluster with a ZFS replication or CEPH for the VM to be HA ready would definitely be possible. Otherwise as @euronodes already said, syncing all these components (sessions, uploads, database) take significant work.

Depending on your requirements, maybe a provider with a really good uptime that also offers SLA (99,99%) like IONOS, could also be beneficial.

euronodes · May 18

@matserix said:
A 3 node Proxmox cluster with a ZFS replication or CEPH for the VM to be HA ready would definitely be possible

Technically of course, but in practice it doesn make a sense for a woo website, unless it's a TEMU clone that justifies the expense.
OTOH I've heard that huge TEMU warehouse burned to the ground with almost half million items iniside.
The total damage was north of €87.99

Chievo · May 18

@matserix said:

@Chievo said:

@euronodes said:
What do you want to HA? Some platform, website, eshop? Strategy depends on the product TBH

Configuration is pretty simple woo+wordpress website. Transactions as well content as well.

Tho not simple to setup high availability.

A 3 node Proxmox cluster with a ZFS replication or CEPH for the VM to be HA ready would definitely be possible. Otherwise as @euronodes already said, syncing all these components (sessions, uploads, database) take significant work.

Depending on your requirements, maybe a provider with a really good uptime that also offers SLA (99,99%) like IONOS, could also be beneficial.

Thanks for your comment yes proxmox Ha should be tried I guess thanks for the advice

Chievo · May 18

@euronodes said:

@matserix said:
A 3 node Proxmox cluster with a ZFS replication or CEPH for the VM to be HA ready would definitely be possible

Technically of course, but in practice it doesn make a sense for a woo website, unless it's a TEMU clone that justifies the expense.
OTOH I've heard that huge TEMU warehouse burned to the ground with almost half million items iniside.
The total damage was north of €87.99

Is not a temu clone . But well I must do the math I guess to check it thanks

slowservers · May 18

How bad is your downtime currently? How much downtime is acceptable? Do you know why it's going down?

Shared hosting isn't necessary unreliable, but it can be.

Is this a static site? Dynamic but readonly, or dynamic and writable?

Chievo · May 19

@slowservers said:
How bad is your downtime currently? How much downtime is acceptable? Do you know why it's going down?

Shared hosting isn't necessary unreliable, but it can be.

Is this a static site? Dynamic but readonly, or dynamic and writable?

I have had some 6h downtime recently in one day . ( The past month 2 days downtimes around 4h each ) Yeah I know it is not critical and for the price is OKish( I am not trying to say that for shared hosting should be better for the cost is OK I am aware that better things cost exponentially more money but well ) but if could be avoided could be better somehow. Less downtimes I mean 6h is not too bad but depends when happens not the same during the night than during the day but it is luck .

It is a woo+ wordpress . Dynamic

WSS · May 19

Move to a shared provider who doesn't incredibly oversell their shared hosting. It's been a loss leader for years, but most just let the box take more clients than it should and don't police the ones already there.

ICDSoft isn't cheap, but I had maybe 8 hours downtime with a decade of them, and that was after they called to tell me they needed to move me between datacenters.

Chievo · May 19

@WSS said:
Move to a shared provider who doesn't incredibly oversell their shared hosting. It's been a loss leader for years, but most just let the box take more clients than it should and don't police the ones already there.

ICDSoft isn't cheap, but I had maybe 8 hours downtime with a decade of them, and that was after they called to tell me they needed to move me between datacenters.

Thanks for the suggestion yeah it seems cost effective option thanks for the suggestion once again !

Falzo · May 19

@slowservers said:
How bad is your downtime currently? How much downtime is acceptable? Do you know why it's going down?

Shared hosting isn't necessary unreliable, but it can be.

Is this a static site? Dynamic but readonly, or dynamic and writable?

these are the right questions!

advising to switch from some just not so reliable cheap shared hosting to a (self-managed?) HA setup is rather overdoing and not accounting for a lot of potential trade offs financially, work-wise and so on.

there is lot of middle grounds here, like simply switching to a more reliable provider like @WSS suggested. most large and well-known provider have very good track record.

matserix · May 19

@Falzo said:

@slowservers said:
How bad is your downtime currently? How much downtime is acceptable? Do you know why it's going down?

Shared hosting isn't necessary unreliable, but it can be.

Is this a static site? Dynamic but readonly, or dynamic and writable?

these are the right questions!

advising to switch from some just not so reliable cheap shared hosting to a (self-managed?) HA setup is rather overdoing and not accounting for a lot of potential trade offs financially, work-wise and so on.

there is lot of middle grounds here, like simply switching to a more reliable provider like @WSS suggested. most large and well-known provider have very good track record.

Yes and no, depending on the loss of something like 6 hours or even more in the rush time, it's definitely worth to have a true HA with a max downtime of seconds to a minute. I think @Chievo should do the maths to see what's best for his use case. Having a complete outage because you are single homed also doesn't feel nice, that's at least how I feel about it.

Chievo · May 19

@Falzo said:

@slowservers said:
How bad is your downtime currently? How much downtime is acceptable? Do you know why it's going down?

Shared hosting isn't necessary unreliable, but it can be.

Is this a static site? Dynamic but readonly, or dynamic and writable?

these are the right questions!

advising to switch from some just not so reliable cheap shared hosting to a (self-managed?) HA setup is rather overdoing and not accounting for a lot of potential trade offs financially, work-wise and so on.

there is lot of middle grounds here, like simply switching to a more reliable provider like @WSS suggested. most large and well-known provider have very good track record.

Thanks for the suggestion ! Yes something in the middle would be the sweet point tbh.

Chievo · May 19

@matserix said:

@Falzo said:

@slowservers said:
How bad is your downtime currently? How much downtime is acceptable? Do you know why it's going down?

Shared hosting isn't necessary unreliable, but it can be.

Is this a static site? Dynamic but readonly, or dynamic and writable?

these are the right questions!

advising to switch from some just not so reliable cheap shared hosting to a (self-managed?) HA setup is rather overdoing and not accounting for a lot of potential trade offs financially, work-wise and so on.

there is lot of middle grounds here, like simply switching to a more reliable provider like @WSS suggested. most large and well-known provider have very good track record.

Yes and no, depending on the loss of something like 6 hours or even more in the rush time, it's definitely worth to have a true HA with a max downtime of seconds to a minute. I think @Chievo should do the maths to see what's best for his use case. Having a complete outage because you are single homed also doesn't feel nice, that's at least how I feel about it.

Yes I agree with you. It is feeling bad during the downtime. Yes doing the math when I study all the costs etc.

Falzo · May 20

@matserix said:

depending on the loss of something like 6 hours or even more in the rush time

absolutely. yet, for a small shop that runs off a cheap shared hosting, this loss will be absolutely negligible. people always tend to see, that they do get not some orders and cry wolf about losing millions. however in reality think about the real user journey.

someone that is close to make a decision of actually buying or even decided to do so, will in most cases not just change their mind and turn away but simply order in the next days then :-)
and people who are just window shopping even more, yeah they can't see your offers for a moment, but eventually come back later. yet they would not have brought in money immediately anyway.

so in reality all you see there is a shift to a later date and the real loss/churn is gonna be far less than what people usually cry about.

I'd still say, any online shop that doesn't make thousands per day, does not need HA in the seconds or minutes.
but of course, reliabilty in general matters. if you have some downtime every other day, that is a no go.

just pick another provider, the bigger ones usually have proper monitoring and some internal failover/HA in place already, so you'll eventually get both without owned risk and work overhead :-)

maybe it really is worth thinking about math aka talking about money. what do you pay for the shared hosting and how much percent is that spending in relation to your earnings @chievo? e.g. if you earn $1000 a month but spend $1 ...

Chievo · May 20

@Falzo said:

@matserix said:

depending on the loss of something like 6 hours or even more in the rush time

absolutely. yet, for a small shop that runs off a cheap shared hosting, this loss will be absolutely negligible. people always tend to see, that they do get not some orders and cry wolf about losing millions. however in reality think about the real user journey.

someone that is close to make a decision of actually buying or even decided to do so, will in most cases not just change their mind and turn away but simply order in the next days then :-)
and people who are just window shopping even more, yeah they can't see your offers for a moment, but eventually come back later. yet they would not have brought in money immediately anyway.

so in reality all you see there is a shift to a later date and the real loss/churn is gonna be far less than what people usually cry about.

I'd still say, any online shop that doesn't make thousands per day, does not need HA in the seconds or minutes.
but of course, reliabilty in general matters. if you have some downtime every other day, that is a no go.

just pick another provider, the bigger ones usually have proper monitoring and some internal failover/HA in place already, so you'll eventually get both without owned risk and work overhead :-)

maybe it really is worth thinking about math aka talking about money. what do you pay for the shared hosting and how much percent is that spending in relation to your earnings @chievo? e.g. if you earn $1000 a month but spend $1 ...

Hi! Well like I have said it is not a critical. I am not losing millions but makes me feel "unprofessional" if every month the web is down for some hours .( It is like everything some people would say look at this person . He is so unprofessional with this. Imagine the rest would be the same. It is like quality of packaging of the products the boxes or cleaning every X hours the shop/clinic or whatever and in front of it - at street level ) ( Could create bad image so it is really BAD in my eyes ) It is like I have not done my homework let s say . Yeah if someone is aware that someone have HA in place and is offering shared hosting without frequent downtimes . It could be cool . But not sure which of them are using it for example . WSS suggest icdsoft but if someone is aware of another providers would be better . You know having more options is always beneficial . If is shared hosting with HA and reliable I am happy to pay more money . Just to not look "unprofessional" .

I do not agree with that . I mean from my perspective is the next . I am ready to buy or get a service at A.com . I am going it is down . I am going to B.com . And probably would stay with B.com from this moment . Never at least in the near future I would check A.com again . May be his service or product is better and just the website but meh . I am staying with B.com . It depends from the product but saving some euro in hosting is a bad deal if is not reliable . You would lose customers and decent money at mid term . At least this is my perspective.

About the math I must calculate the HA costs would be absolutely higher but may be better and reliable option . This alternative must be studied so I would not say neither is viable or not. Just depending of the math and well even the math is going bad may be I would sleep better . If is really really bad math may be not a viable option. But let s see .

I am using shared hosting . It is cheap. Around 6 euros per month but I have not selected it because the price tag . I have had a good experience before not sure if is bad luck or the provider is overselling it more . Or the service is going to worse . Like I have said I do not need to "save money" on shared hostings . It is more for image and reliability and sleeping better knowing that my homework is done.

angstrom · May 20

@Chievo said: I have had some 6h downtime recently in one day . ( The past month 2 days downtimes around 4h each )

Wow, that's not great even for inexpensive shared hosting

I would echo the sentiment that you should first invest in a more performative/reliable shared hosting plan

Chievo · May 20

@angstrom said:

@Chievo said: I have had some 6h downtime recently in one day . ( The past month 2 days downtimes around 4h each )

Wow, that's not great even for inexpensive shared hosting

I would echo the sentiment that you should first invest in a more performative/reliable shared hosting plan

Thanks I am going to wait some days and move it to another shared provider ( icdsoft or another one that is recommended here ) meanwhile studying the costs for the HA. But getting covered you know

Falzo · May 20

depending on the ressources and location you need, just go with one of big ones. even hetzner or OVH offer webhosting plans.
but maybe some who are specialised in shared hosting are even better.
plan on spending 10 bucks a month at least or so. I can recommend some german providers with a good track record that are not neccessarily in the lowend sector but still affordable (think of all-inkl.com, netcup, ionos)

euronodes · May 20

@Chievo

Well, we have a client who is the only one that had few downtimes a day. Nobody else had.
Finally we added him €200 free credit to his account and bumped the VM to 8 cores 32GB to see if this will change anything (enough for few months testing)

So far no downtime. He was running seemingly very light website on 2 cores, 4GB RAM but on top of cPanel on top of Alma with CPU hitting 120% to the point even NoVNC console crawled

So, you know, maybe it's not necessarily webhosting company, coz those downtimes for few hours sound a bit absurd

Chievo · May 20

@euronodes said:
@Chievo

Well, we have a client who is the only one that had few downtimes a day. Nobody else had.
Finally we added him €200 free credit to his account and bumped the VM to 8 cores 32GB to see if this will change anything (enough for few months testing)

So far no downtime. He was running seemingly very light website on 2 cores, 4GB RAM but on top of cPanel on top of Alma with CPU hitting 120% to the point even NoVNC console crawled

So, you know, maybe it's not necessarily webhosting company, coz those downtimes for few hours sound a bit absurd

Thanks for the suggestion well I would move to another provider meanwhile checking another alternatives .

Chievo · May 20

@Falzo said:
depending on the ressources and location you need, just go with one of big ones. even hetzner or OVH offer webhosting plans.
but maybe some who are specialised in shared hosting are even better.
plan on spending 10 bucks a month at least or so. I can recommend some german providers with a good track record that are not neccessarily in the lowend sector but still affordable (think of all-inkl.com, netcup, ionos)

Thanks all-inkl shared plans seems OK . ( Over 10 euros border and limited to 50 users per server. Thanks !

How to achieve Less downtimes ?

Comments