How-to
Copy first, then cut over
Copy the site once while it stays live, check it, then switch over with a short freeze. The final sync copies only what changed since the copy, so the freeze takes minutes, not hours.
- CopyA verified migration while the site stays live
- Before cutoverThe jobs the new site needs, and an optional nightly sync
- Freeze the sourceStop changes, then check nothing changes
- Final syncOnly what changed since the copy
- Move the domainInsites Console, then DNS
- After cutoverThe jobs that need the domain
- Watch for 14 daysEvery page, and the log, every six hours
- CloseClose the migration; the report stays
The launch, in time order
The Launch stage opens with a timeline of the whole launch, and marks where you are: Before cutover, Cutover, After cutover, Watch and Close, each a step of its own in that order. Under the timeline, a table says what happens in each of the four steps you work through, Before cutover, Cutover, After cutover and Watch, and what each one does for you. The row for the step you are at is marked. Only this stage shows the timeline.
A step of the launch shows as done, a filled dot with a tick, only when every job or item in it is done or deliberately dismissed. Where you are is the first step that still has something open, so marking a cutover step early does not move you past a job still open before cutover. Under each step a short count says where it stands, for example "6 of 7 done" or "Not started", and the line between the dots fills only up to where you are. Ticking the last open job of a step marks it done and moves you on at once. The same rule applies to the list of stages on the left: Verify gets its tick only once nothing is open under Things that need you, and until then it shows how many, for example "Needs you: 1".
Each job on the To do list sits in the step it is done in. Before cutover holds what the new site needs from its first visitor, and can be done while the old site is still live: keys and integrations, reCAPTCHA among them, constants, forms and their emails, code and settings. The Cutover step holds the jobs that move the domain. After cutover holds what needs the domain on the new site: telling people the site has moved, switching webhooks back on, telling search engines, and deleting the old site.
Webhooks are copied switched off, so a copy cannot send a real message. The job Switch webhooks back on after cutover lists each webhook that was on on the old instance, by its event and the host it sends to. Switch on only those: a webhook that was off on the old instance stays off. If no webhook was on, there is no job.
On a move from an Insites instance to v6, Before cutover can also hold jobs about the copied code. Each lists every file and line it means. Fix the links that still depend on the old site lists each link that still names the old instance, or a file address that does not open (the LINKS check). Change the names in the copied code lists each old module name the migration could not change safely, and each short name of a query, form or database it could not make full, with the full names that end in it. Review the pages that differ lists only pages that were compared and differ, never a page the old site could not answer, and ends with Check again, so you can see which pages still differ once you have fixed them. A job that says to press Check again has the button beside it.
Keep the copy close with a nightly sync
When the first load ends, the Nightly sync step of Load asks whether to keep the new site in step with the old one every night until cutover. Choose Sync every night or Do not sync. A sync you choose there starts as soon as the load is verified. The same choice is under Before cutover in Launch, where you can change it. Each night, in the early hours Sydney time, it reads only what changed on the old site since the last sync, loads it, and checks it. You get one email after each night's sync with its result. A night whose changes need a decision, such as a row edited on both sites, does not load: it emails you to decide. It runs for 7 nights unless you stop it, and switches itself off when you mark the final sync or the domain move done, because from then on people edit the new site. The tool keeps both keys while it is on.
When a night waits for you. When someone changed rows on the new site after they were copied, copying that night would write over their changes, so the night stops before copying anything. The migration's row on the Migrations list and the top of the migration say Nightly sync is waiting for you, and the email names each table. Open it, and Load lists each table with how many rows changed. For each one choose Keep the new site's rows, which leaves those rows as they are and stops the nightly sync asking about that table again, or Take the old site's rows, which copies them again from the old site and loses the changes. Then press Carry on to load the night. A night that stopped because another migration was working on the new site asks nothing: press Carry on once that migration has finished.
Code changed on the new site is kept. Every sync, the nightly sync, the final sync and Run the final sync again alike, leaves a code file as the new site has it when someone changed it there after the migration last copied it: a page, layout, partial, email or text message notification, API call, form, sign-in rule or GraphQL query. The sync's log says kept as the new site has it with the file, the report lists it, and the To do list adds Review files of code kept as the new site has them, so you can check whether the old site's own later change to that file also needs making on the new site. A file the old site added is still copied, and a file nobody changed on the new site still takes the old site's newer version. When the tool cannot tell whether a file was changed on the new site, for example on a migration loaded before this change whose earlier copy cannot be read, it leaves the file alone and lists it the same way, rather than writing over it.
A stylesheet changed on the new site is kept too. A website's stylesheet that the tool rewrote on the way across, such as a Webflow site's shared stylesheet under assets/webflow/css/, is treated the same way. If someone changed it on the new site after the migration last copied it, for example by adding a CSS fix at the end, the sync leaves it as the new site has it, and every page's link to it is given the new file's integrity hash so the browser still loads it. The log says kept as the new site has it with the file, and it is listed in the same To do. If the old site's stylesheet also changed since, make that change on the new site yourself. Images, fonts and scripts the new site already holds are never copied over again by a sync.
When a key stops working. Before each night the tool checks both keys still work. If one does not, for example because a newer token was taken in Insites Console, that night is skipped, and the migration's row on the Migrations list and the nightly sync on Before cutover both say Nightly sync paused: the new site's key no longer works (or the old site's), with Give the migration a new key. It runs again the next night once the key works.
Each night's sync sits under its migration: on the Migrations list it is a Nightly sync under Show later syncs, started by The nightly sync, and on Activity its lines name the migration with the night's sync beneath it.
What changed since the last sync. A final sync and each night of the nightly sync check everything again. Their Verify compares each check with the run before it, the first load or the sync before, and says how many changed: "Changed since the last sync: 2 of 26 checks". Each of those is marked Changed since the last sync on Things that need you and in the Verification log, with whether it passed last time, how many items fail now that did not then, and how many that failed then pass now. A check that did not run last time is marked New since the last sync. A count that only grew with the site, such as 1,203 of 1,203 records and then 1,240 of 1,240, is not a change. When nothing changed, Result says so.
The cutover
If any check still fails, or you accepted something and have not checked again, the Cutover step of Launch lists what fails and holds back the final sync and the domain steps. Fix or accept each one on the Things that need you step and check again. If it is fine to go ahead anyway, say why under Go ahead anyway: it is asked once, and the reason goes in the report. A go-ahead also settles those failures on Result: the check stops counting under Not confirmed and Things that need you, the migration reads as Verified, and the reason is kept with the check. It stays that way through Check everything again. If the check later fails for something the go-ahead did not cover, it asks again.
When the reason and the To do list disagree. If your reason says you skipped something, such as "I skipped the domain and DNS settings", while a job about it is ticked as done, the tool asks Are you sure? and names the ticked jobs. Press Cancel and untick them, so the list says what is really done, or go ahead: the report keeps the reason and the jobs it contradicts.
- Freeze the source. Tell people, or put it into maintenance. Check for changes shows how many records changed since the copy read the source.
- Final sync. Everything happens on this step. The sync compares the old site with the ledger the load kept of every record it copied, and says what it found, for example Will copy 3 new, update 2 changed, 0 deleted on the old site. Look through them lists each database. Type the plan code and press Load: it copies the new records, updates each changed one where the first load put it, and then shows a short check of what it changed. A record deleted on the old site stays on the new site. Each count that is not 0 opens to its records, by database and old-site id, so you can find the record on either site. When the sync is done, its line names each check that did not pass, with a link to it in the Verification log. A code file the sync kept as the new site has it, because someone changed it there, says so in the check: if that change was on purpose, press Keep the new site's version to accept it. A record the first load copied that the tool cannot find on the new site is listed and never copied without you, and a sync that would create more records than are new stops and says why. A sync planned by a version before 0.8.13 shows Count the changes again instead of Load. Its Transform choices are the migration's own, so it never opens Transform. To read only what changed, the sync uses the copy of the old site the tool kept from the run before it. That copy is removed 7 days after the run's last activity, so a final sync more than a week after the copy reads the whole old site again: it takes longer, and it still loads only what is new or changed since the last load. A final sync that stopped with The final sync needs to read the whole old site wrote nothing: press Carry on. Once it has run, the button becomes Run the final sync again, for the rare case the old site changed after it. It never writes over a code file you kept as the new site has it: see Keep the new site's version. It needs both keys: once the copy is verified nothing keeps them unless the nightly sync is on, so after a day or more it may ask you to give the migration a new key first. How long a key is kept.
- Move the domain. Add it to the new instance in Insites Console, then change its DNS records. If you add the domain in Console before its DNS points at the new instance, Insites emails you "Domain couldn't connect: your domain", marked Failed. That is expected at this stage, and you do not need to email support. Once the DNS has moved, Console connects the domain, and you use Check again on the job. The job on the Cutover step says this with your domain's name in it. The jobs under the four steps name the domain and the records to change. When you tick a domain or DNS job, the tool opens each domain it names, as a visitor would, and checks it answers from the new instance. If it still answers from the old site, for example from Webflow, the job stays ticked and says Ticked, not yet true, with what the domain answered from. That is not an error: it is expected until the domain points at the new site, and a domain that cannot be reached yet says so the same way. Activity records the tick the same way. DNS can take a while to spread: press Check again on the job once it has, and the job says when it became true. You do not have to come back to tick them. Each time you press Check again under Checks on the new site in Launch, and once a day while the tool watches the new site, it opens every domain an open domain or DNS job names. When they all answer from the new instance, it marks the job done by the migration tool, and Activity records it.
- Turn on emails and connections. If the plan held the site's emails and outside connections until cutover, press Turn on emails and connections on Cutover once the domain points at the new site, or let the domain step do it if you chose that. See Test on a live instance safely.
- Do the After cutover jobs. They are on the After cutover step: each needs the domain on the new site.
- Watch the new site. On the Watch step, Start the 14-day watch as soon as the domain points at the new instance. Starting it takes a moment, and you can leave the page while it starts: the watch runs on the tool, not in your browser. The first look, and the first email, come within about 5 minutes. Every 6 hours for 14 days, 56 times in all, it opens each page the verification checked, as a visitor, and compares the answer with the one at verification, and reads the new instance's log for new errors. It changes nothing. When the new instance logs an error that names no file while the watch is opening pages, such as
Could not find Table with name: modules/logs/form_debug_log, the watch names the pages it opened at that time. Such an error may come from an installed module that runs when one of those pages opens, and any visitor would cause it too, so it is still counted. When it asks for a table the instance does not have, the watch says whether an installed module has that folder. A page that asked visitors to sign in when the migration was verified, and now opens without it, is flagged as high priority: the watch says the page no longer asks visitors to sign in, and tells you to open it in a private window and check the page's authorization policy and its layout. You get one summary email a day, 14 in all, one when it ends, and one straight away the first time something goes wrong. The tool keeps the new instance's key until you close the migration, so the watch reads the new site's error log for all 14 days. To have each problem filed as a task in Teamwork too, choose a task list on the Watch step: see file the watch's problems in Teamwork. To stop the watch and its emails sooner, open the Watch step and press Stop watching. Each watch email says this and links straight to the step. You need the operator role, and Activity records who stopped it. - Close. When the new site has run cleanly, press Close the migration on the Close step and type the old site's address, or its Insites name, which the dialog shows. An old Insites instance is shown by its Insites name, such as
acme-v5, and typing the name or the address both confirm. If the 14-day watch or the nightly sync is still on, the dialog says first that closing ends it, how long ago the watch started and when it was due to end, and which checks still fail, and you tick a box to say you understand before it closes. The record of the close says what it ended. The dialog then lists anything else still open, such as a check that did not pass or a cutover step not marked done. Nothing on either site changes, and the tool removes its copies of the keys. Closing waits while any run of the migration is still working, such as a final sync that is loading, and says which. After that the migration can no longer be rolled back, checked again or synced, and its report, its decisions and its load ledger stay. Deleting the old instance is a job under After cutover. - Close only a live site. Close is only for a site that is live on its domain. A closed migration gets no more checks, copies or syncs until an owner reopens it. Until the Move the domain step on Cutover is marked done, the tool refuses to close and says so. An owner can tick Close anyway and say why, and the reason is recorded with the close. Operators cannot close anyway: they mark the domain step done once the site is live, or ask an owner.
- Reopen. An owner can reopen a closed migration from the Close step with Reopen the migration, typing REOPEN and saying why, for example to bring a fix from a later version across to the launched site. The dialog says what reopening allows: Check again, Copy all again, a final sync and the nightly sync, each of which writes to the live site. Every decision, every acceptance and the load ledger stay, so a final sync copies only new or changed rows, and code and pages as the current version makes them. A fix made on the new site since is never written over unless you choose to, item by item. Close removed the keys, so give the migration a new key for each site after reopening. Who reopened it and why are recorded in its decisions, which the report and the Activity page list. Close it again when you are done. Operators and viewers do not see Reopen.
File the watch's problems in Teamwork
The 14-day watch can file what it finds as tasks in Teamwork, so a problem lands where your team already works rather than only in an email. It is optional: without it, the watch emails only, as before.
- Choose where. On the Watch step, under File problems in Teamwork, press Choose a Teamwork task list. Enter your Teamwork site and API key, or use your saved key, then choose the project, the task list, and who the tasks are assigned to. The person choosing is picked by default when they are on the project; choose Nobody to leave the tasks unassigned. The dialog always asks: it does not take the task list the To do list was pushed into. Do it before you press Start the 14-day watch, or at any time while the watch runs. Press Change to choose another task list, or Stop filing to go back to email only. Tasks already filed stay in Teamwork.
- One task per problem. Each distinct problem on your site becomes one task: a page that now fails, sends visitors somewhere else, shows an error or does not answer, or a new error in your site's code in the new instance's log. The task is named after the site and the problem, and says what was seen and when, the page's address, what the page answered when the migration was verified, and links back to the watch in the tool.
- Seen again. A problem the watch sees again adds a comment to the same task, never a second task. A problem that carries on adds at most one comment a day, not one at every 6-hour look.
- Seen fixed. When a page answers as it did at verification again, the watch adds a comment saying so. It never completes the task: that is for your team. If the problem comes back, the watch comments on the same task again. An error in the log is never said to be fixed, because the log only shows what is new, and an error not logged again has not shown it is gone.
What is not filed. An error in code an Insites module provides is listed in the email, with nothing for you to fix, and is not filed. Nor is the watch's own trouble, such as a page it did not reach in time.
The emails. Each watch email says which project and task list tasks go to, and who they are assigned to. The first email about a problem, and each daily summary, link the task filed for each problem. Without a Teamwork task list chosen, an email that found problems says that choosing one on the Watch step files them as tasks automatically. If a task cannot be filed, for example because Teamwork is down or the saved key no longer works, the email still goes, says the task could not be filed and why, and the watch tries again at its next look.
The key. The watch files tasks on its own, every 6 hours, so it uses the saved Teamwork key of the person who chose the task list, only for the Teamwork site it was saved for. A key typed in this dialog is always saved; see your saved Teamwork key. The watch's own use of the key does not count as use, so a key used only by the watch is still removed after 180 days without a person using it. Teamwork gets the problem, the page's address and the link back. It never gets a key, a password or any of the site's data. The same addresses work as for the push: see which addresses work. If the migration is moved to an agency, the choice moves with it, and filing carries on once that person has saved their Teamwork key in the agency.
Push the To do list to your project management tool
The jobs on the To do list are the ones only a person can do, such as moving the domain and setting the constants that were not copied. To work through them where your team already works, press Push to your project management tool on Before cutover or on After cutover, in Launch, then choose Teamwork. Monday.com and ClickUp are coming soon.
Each of the two pushes its own jobs. Before cutover pushes the jobs before the cutover and the ones at it, such as moving the domain. After cutover pushes the jobs that need the domain on the new site, and names each task After cutover, so your team can see it waits for the domain to move.
- Your Teamwork site and key. Enter the address you open Teamwork at, such as
https://yourcompany.teamwork.com, and your Teamwork API key. In Teamwork it is under your profile picture, Edit my details, API and Mobile. Leave Remember this key ticked and you will not enter it again: next time the dialog starts with your saved key in use. Press Use a different key to enter another. - Choose the project, then the task list. The tool loads your projects, then the task lists of the one you choose. Once a push has been made, the dialog offers its site, project and task list first, so both screens' jobs can go into one task list. Press Choose a different task list to pick another.
- Choose a task to put them under, if you like. Once a task list is chosen, Under task lists every open task in it, at any depth: tasks, their subtasks and the subtasks below those, each indented, with the tasks above it named under it. Search finds a task by its own name or by the name of any task above it. Leave it at None to add the jobs at the top of the task list, or choose a task to add them as its subtasks.
- Push. Each of that screen's jobs that is not done yet becomes one task in that task list, or one subtask of the task you chose: the job's title, why this site needs it, its steps, and a link back to the job in the tool. They go in the order of the launch. The dialog then lists each task with a link to it in Teamwork, and the screen shows how many of its jobs were pushed, and where.
Push again whenever you like. A job already in that task list, under any task or none, is not added twice, even if someone has completed its task in Teamwork; the dialog says how many were created and how many were already there. A task deleted in Teamwork is added again. Pushing into a different task list adds them all there.
The sync is two-way. Each pushed job stays linked to its task. Mark the job done in the tool and its task is completed in Teamwork; reopen the job and the task is reopened. Complete or reopen the task in Teamwork and the job follows in the tool, straight away when Teamwork lets the tool watch the project, and otherwise at the tool's regular check, within about 30 minutes. Each change is listed in Activity, saying which side it came from. Changes made in the tool are sent with a saved Teamwork key: yours if you have one for that site, otherwise the key of the person who pushed, so leave Remember this key ticked. Comments are not copied either way. Jobs pushed before version 0.8.4 are linked once you push again into the same task list.
What is sent, and what is kept. Teamwork gets each job's title, why, steps and the link back, and the two site addresses of the migration. It never gets a key of either instance, a password or any of the site's data. Your Teamwork API key goes from your browser to the tool over an encrypted connection. It is never written to a log, or put in the migration, its reports, its emails or anything sent to AI. The migration records which task list the jobs went to, each task's number, and who pushed them and when, and the technical report lists each push.
Your saved Teamwork key. With Remember this key ticked, the tool keeps your key once Teamwork has accepted it, encrypted like the instance keys. It is yours alone: it is kept for you in the account you are working in (Personal, or one agency), only you can use it, and it is only ever sent to the Teamwork site you saved it for. It is never shown again, not even to you; the dialog says only which site it is for. Settings, in the menu on the left, shows its site and when you saved it. It goes when you press Forget my Teamwork key, on Settings or on the push dialog; when you are removed from the agency; and when you have not used it for 180 days. If Teamwork stops accepting it, for example because you made a new key in Teamwork, the dialog says so and asks for a new one, which replaces it. Saving, using and forgetting it are recorded on Activity.
Which addresses work. The tool sends your key only to a Teamwork address: https, ending in teamwork.com, such as https://yourcompany.teamwork.com or https://yourcompany.eu.teamwork.com. Your own custom Teamwork domain, such as https://pm.youragency.com, works too, with nothing to ask for: before sending your key, the tool checks that the domain's DNS has the CNAME record Teamwork asks for when you set up a custom domain, pointing to sslproxy.teamwork.com (or sslproxy.eu.teamwork.com for an EU account). An address that only looks like Teamwork fails that check and gets no key. The tool does not follow a redirect.