Slow Azure Image Recovery
  • Davanne
    0
    Hi,

    I have been running DR test to azure, the restore process is very slow.

    restoring a 50GB server took 1.5 hours to complete, the Azure instance is running on SSD and has 1GBPS connection. However the Cloudberry application seems to limit the transfer to 70mbps-85mbps. Not sure how i can possible use this as a DR solution with such slow restoration. I have logged few tickets regarding this but it's dead end and I am starting to look at different solutions. But if someone from Cloudberry can address this with urgency please !!.



    Also any one else can share their feedback when restoring a Baremetal server to Azure or AWS , what is your average RTO and what size data.
    Davanne Options
  • Denis Gorbachev
    6
    ↪Davanne

    Hi. Please, PM me your ticket numbers, so I could check.
    Denis Gorbachev Options
  • Davanne
    0
    sent you the ticket numbers
    Davanne Options
  • Davanne
    0
    Hi Denis,

    I have sent email to support but no response, I asked what was the average restore time you got when restoring to Azure during your testing but no response for a week now from Vlad
    Davanne Options
  • Denis Gorbachev
    6
    ↪Davanne
    I've notified the support team. They will get back to you shortly.
    Denis Gorbachev Options
  • Davanne
    0
    Hi Denis, do you know if there is any update on this?
    Davanne Options
  • David Gugick
    118
    ↪Davanne
    Support has the case and someone will be reaching out to you shorty to provide an update. Thanks.
    David Gugick Options
  • James Dyke
    4
    I have also tested and failed. the problem you have is the data has to be "copied" so this is why the recovery is slow. there is no other way around it, Veeam also seem to do this the same way.

    once the VM is booted they then need to copy all the data into the vhd's the live server can then be spun up. like you say very time consuming and not good in a BC/DR scenario.
    James Dyke Options
  • Davanne
    0
    Yep, logged a ticket about 2 months ago no response
    Davanne Options
  • Julia
    4
    ↪Davanne
    Thanks for bringing it up this time! As for restoring to AWS there is a known bottleneck that creating an instance can take a long time - it is under the AWS itself and we can’t control that.
    As for Azure, it seems to be something that we should go deep into. I am going to review the customers complaints (with diagnostic logs) and double check with our team what speed scores we have and share in this thread.
    Julia Options
  • Davanne
    0
    Yeah we where force to look into other products, its a shame as if this worked well then it would have been a great DR solution that is affordable and easy to sell
    Davanne Options
  • Julia
    4
    ↪James Dyke
    Appreciated for your comments!
    I reviewed your tickets as well - the most recent one (50093) is that you had some troubles with Azure VM booting on restore, was it finally resolved? There is no update in the support ticket from you.
    And the only concern now is a slow restore to Azure as @Davanne faced with, right?
    Julia Options
  • Davanne
    0
    That's correct, the restore is really slow... if we could get this down to 4hours for 800-1TB of data that would be super... But right now it is 50GB per 1.5 hours. Our clients have data 500+ GB it would take 20hours+ to get them back online.
    Davanne Options
  • Julia
    4
    ↪Davanne
    Please find some numbers that I was able to collect from my review:
    - our internal test shows 250GB per 3-4h that is ~100GB per 1.5h.
    - another customer reported the following:
    "I am running the “TEST” on creating the RESTORE right now . . . but I would Imagine it will take several hours to create a 380 GB VHDX image of the server. Right now it is taking about 35 minutes to restore 10%. Guessing that it will take about 6 hours to create an IMAGE that is in the VHDX format".
    That means that they had similar speed as ours (100GB - 1,5h).
    Julia Options
  • Davanne
    0
    You will need to cut it down to atleast 250-300 GB per 1hour to make this a feasible solution, otherwise its just something on paper to market no value
    Davanne Options
  • Julia
    4
    ↪Davanne
    We have been investigating the case more carefully and I believe we can address some improvements within the next version. I will let you know our results!
    Julia Options
  • Davanne
    0
    Do you have any ETA? as we are looking to move to Solarwinds for backup solution.. If this is improved it will definitely be something that will allow us to stick to Cloudberry
    Davanne Options
  • Julia
    4
    ↪Davanne
    I see... I will be able to give an update by the end of the week.
    Julia Options
  • Davanne
    0
    Thanks
    Davanne Options
  • Julia
    4
    ↪Davanne
    A quick update so far: we were not able to reproduce your slowness, however, we made some changes and going to run a new test to get new results next week. It is coming to be v5.9.
    Julia Options
  • Davanne
    0
    Whats speeds did you get?
    Davanne Options
  • Julia
    4
    ↪Davanne
    The new test is coming, no speed results yet. Sorry.
    Julia Options
  • Julia
    4
    ↪Davanne
    A quick update - I have to say that we got almost the same speed: 100GB - 1,5h. This is a continuous process, no timeouts or interruptions there.
    Julia Options
  • Davanne
    0
    Well when you include encryption and compression that will increase to about 2hours+
    Davanne Options
  • Julia
    4
    ↪Davanne
    That’s true... it takes 2-2,5h per 100GB with compression/encryption. Is your main use case a restore to Azure?
    Julia Options
  • Davanne
    0
    Yes, its a really affordable DR solution for small businesses. We can offer it with a good pricing which make's it very easy to sell to our clients. But we can't have restore take 12-24 hours to complete.
    Davanne Options
bold
italic
underline
strike
code
quote
ulist
image
url
mention
reveal
youtube
tweet
Add a Comment