A large import, backup or image task can run out of PHP memory even while ordinary pages continue to work. Restarting the entire job may repeat the same failure and can make partial results harder to understand.
First record what completed successfully, then test a representative smaller batch. That helps you distinguish an unusually heavy item from a workload that simply needs smaller batches or a verified increase in available resources.
This guide separates job recovery from the general memory-limit error. Preserve the source and any known usable backup so that retries do not create duplicates or replace the only good copy.
Guide reviewed: October 8, 2026. Interface labels may vary by version, account and device.
Preserve completed work before retrying the job
| Job pattern | Compare | Next check |
|---|---|---|
| Import fails partway | Small representative batch | Resume behavior and duplicate handling |
| Same record always fails | That record in a small sample | Malformed or unusually large data |
| Backup grows unexpectedly | Selected scope and exclusions | Old archives included in the new job |
| Image fails despite modest file size | Pixel dimensions and format | Decoded memory and image library |
| Several jobs overlap | Resource graphs at the failure time | Scheduling and account capacity |
Establish what completed before the failure
- Save the full error and failure time.
- Check the job log for the last completed item.
- Inspect partial output without treating it as complete.
- Confirm whether the tool resumes or starts again.
- Use a staging copy or safe sample before testing a retry.
Look at the job’s own log and output before clicking Start again. An importer may already have created records; a backup may have written a partial archive; an image tool may have processed some sizes but not others.
Keep the last completed item, error time, and full PHP message. The “tried to allocate” value is only the final allocation that failed, not the complete memory requirement. A tiny value there does not mean the task needed only a tiny adjustment.
Choose a safe test sample or staging copy. That lets you compare the workflow without creating duplicate products, overwriting a useful backup, or repeatedly processing the entire media library.
For imports, reduce the batch rather than the useful data
Keep a run sheet containing batch size, representative row types, active PHP limit and last successful item. Compare one setting at a time. If the same item fails in every batch size, investigate that item before raising the allowance.
A tool that loads a whole file and builds all its records in memory may need a different processing mode from one that saves small batches. Look for supported batch-size, resumable import, or background-processing options.
Start with a small representative sample that includes the same kinds of fields and images as the full import. If it works, increase the batch gradually within the tool’s recommendations. Record the row count and active memory limit so the comparison is meaningful.
If the same record fails even in a small sample, inspect that record. An unusually large field, malformed input, or repeated image lookup can require a targeted repair. Splitting every file into smaller pieces will not necessarily solve one defective item.
For backups, review what is being collected
Confirm the selected folders and database scope. Old local archives can be included in a later backup if the tool’s exclusions are not configured, causing the job to gather more data than intended.
Keep a known usable backup before removing older copies. A file named “backup” is not enough evidence that restoration will work. Check the tool’s verification options and the storage destination.
Look for chunked archives or resumable transfers when supported. Those methods can make the job more resilient than one browser request that gathers and transfers everything together. The backup tool’s own documentation should determine the supported approach.
For images, compare dimensions as well as file size
A compressed photograph can be a modest upload while requiring much more memory when decoded. The pixel dimensions and the image-processing library affect that work. Compare a suitably resized copy with the original as a controlled test.
Keep the original outside the test before changing dimensions. Check all required generated sizes and the image’s appearance in the actual layout. A successful upload with missing thumbnails is only a partial result.
If one format fails while another works, give the host or developer the format, dimensions, and matching error. That is more useful than increasing the allowance repeatedly without checking the image-processing support.
Check the effective allowance and shared capacity
Review the active PHP memory limit in Site Health and the hosting panel. WordPress memory constants can request changes in supported contexts, but the host may enforce a limit. Confirm the active result rather than assuming a configuration edit took effect.
Several jobs can run at the same time. Scheduling backups, imports, and image optimization together may create pressure even when each one works alone. Compare the account’s resource use with the failure time and avoid unnecessary overlap.
A task that keeps growing without progress needs code or workflow investigation. More memory may delay that failure, but it does not explain it. Ask the developer to review the smallest sample that still reproduces the problem.
Keep the successful workflow
Verify the finished records, usable archive, or generated images. Then keep the batch size, exclusions, and timing that worked. If the same job later fails, those notes give you a clear baseline.
WordPress’s PHP guidance explains its memory settings. The execution-time guide covers another limit that can affect these jobs after memory use is brought under control.
Verify the completed output
For an import, check counts, key fields and duplicate handling. For a backup, use the tool’s integrity and restore-verification procedure on a suitable test environment. For images, inspect required sizes in the public layout. A job reporting completion is useful evidence, but its output still needs checking.